diff --git a/feed_weekly.xml b/feed_weekly.xml index 50ac4cd..3d65496 100644 --- a/feed_weekly.xml +++ b/feed_weekly.xml @@ -7,741 +7,741 @@ http://www.rssboard.org/rss-specification python-feedgen en - Mon, 13 Apr 2026 00:22:35 +0000 + Mon, 20 Apr 2026 00:23:05 +0000 - Self-Execution Simulation Improves Coding Models - https://arxiv.org/abs/2604.03253 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03253.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Gallil Maimon, Ori Yoran, Felix Kreuk, Michael Hassid, Gal Cohen, Pierre Chambon, Yossi Adi</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> A promising research direction in enabling LLMs to generate consistently correct code involves addressing their inability to properly estimate program execution, particularly for code they generate. In this work, we demonstrate that Code LLMs can be trained to simulate program execution in a step-by-step manner and that this capability can be leveraged to improve competitive programming performance. Our approach combines supervised fine-tuning on natural language execution traces, textual explanations grounded in true execution, with reinforcement learning using verifiable rewards. We introduce two complementary objectives: output prediction given code and inputs, and solving competitive programming tasks with either ground-truth or self-predicted execution feedback. These objectives enable models to perform self-verification over multiple candidate solutions, and iterative self-fixing by simulating test execution. Across multiple competitive programming benchmarks, our method yields consistent improvements over standard reasoning approaches. We further present ablations and analysis to elucidate the role of execution simulation and its limitations.</p> - https://arxiv.org/abs/2604.03253 - Wed, 11 Mar 2026 00:00:00 +0000 + SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decoding + https://arxiv.org/abs/2604.09557 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09557.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Talor Abramovich, Maor Ashkenazi, Carl, Putterman, Benjamin Chislett, Tiyasa Mitra, Bita Darvish Rouhani, Ran Zilberstein, Yonatan Geifman</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Speculative Decoding (SD) has emerged as a critical technique for accelerating Large Language Model (LLM) inference. Unlike deterministic system optimizations, SD performance is inherently data-dependent, meaning that diverse and representative workloads are essential for accurately measuring its effectiveness. Existing benchmarks suffer from limited task diversity, inadequate support for throughput-oriented evaluation, and a reliance on high-level implementations that fail to reflect production environments. To address this, we introduce SPEED-Bench, a comprehensive suite designed to standardize SD evaluation across diverse semantic domains and realistic serving regimes. SPEED-Bench offers a carefully curated Qualitative data split, selected by prioritizing semantic diversity across the data samples. Additionally, it includes a Throughput data split, allowing speedup evaluation across a range of concurrencies, from latency-sensitive low-batch settings to throughput-oriented high-load scenarios. By integrating with production engines like vLLM and TensorRT-LLM, SPEED-Bench allows practitioners to analyze system behaviors often masked by other benchmarks. We highlight this by quantifying how synthetic inputs overestimate real-world throughput, identifying batch-size dependent optimal draft lengths and biases in low-diversity data, and analyzing the caveats of vocabulary pruning in state-of-the-art drafters. We release SPEED-Bench to establish a unified evaluation standard for practical comparisons of SD algorithms.</p> + https://arxiv.org/abs/2604.09557 + Tue, 10 Feb 2026 00:00:00 +0000 - SEVerA: Verified Synthesis of Self-Evolving Agents - https://arxiv.org/abs/2603.25111 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2603.25111.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Debangshu Banerjee, Changming Xu, Gagandeep Singh</p><p><b>Upvotes:</b> 27</p><p><b>Summary:</b> Recent advances have shown the effectiveness of self-evolving LLM agents on tasks such as program repair and scientific discovery. In this paradigm, a planner LLM synthesizes an agent program that invokes parametric models, including LLMs, which are then tuned per task to improve performance. However, existing self-evolving agent frameworks provide no formal guarantees of safety or correctness. Because such programs are often executed autonomously on unseen inputs, this lack of guarantees raises reliability and security concerns. We formulate agentic code generation as a constrained learning problem, combining hard formal specifications with soft objectives capturing task utility. We introduce Formally Guarded Generative Models (FGGM), which allow the planner LLM to specify a formal output contract for each generative model call using first-order logic. Each FGGM call wraps the underlying model in a rejection sampler with a verified fallback, ensuring every returned output satisfies the contract for any input and parameter setting. Building on FGGM, we present SEVerA (Self-Evolving Verified Agents), a three-stage framework: Search synthesizes candidate parametric programs containing FGGM calls; Verification proves correctness with respect to hard constraints for all parameter values, reducing the problem to unconstrained learning; and Learning applies scalable gradient-based optimization, including GRPO-style fine-tuning, to improve the soft objective while preserving correctness. We evaluate SEVerA on Dafny program verification, symbolic math synthesis, and policy-compliant agentic tool use (τ^2-bench). Across tasks, SEVerA achieves zero constraint violations while improving performance over unconstrained and SOTA baselines, showing that formal behavioral constraints not only guarantee correctness but also steer synthesis toward higher-quality agents.</p> - https://arxiv.org/abs/2603.25111 - Thu, 26 Mar 2026 00:00:00 +0000 + Turing Test on Screen: A Benchmark for Mobile GUI Agent Humanization + https://arxiv.org/abs/2604.09574 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09574.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiachen Zhu, Lingyu Yang, Rong Shan, Congmin Zheng, Zeyu Zheng, Weiwen Liu, Yong Yu, Weinan Zhang, Jianghao Lin</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> The rise of autonomous GUI agents has triggered adversarial countermeasures from digital platforms, yet existing research prioritizes utility and robustness over the critical dimension of anti-detection. We argue that for agents to survive in human-centric ecosystems, they must evolve Humanization capabilities. We introduce the ``Turing Test on Screen,'' formally modeling the interaction as a MinMax optimization problem between a detector and an agent aiming to minimize behavioral divergence. We then collect a new high-fidelity dataset of mobile touch dynamics, and conduct our analysis that vanilla LMM-based agents are easily detectable due to unnatural kinematics. Consequently, we establish the Agent Humanization Benchmark (AHB) and detection metrics to quantify the trade-off between imitability and utility. Finally, we propose methods ranging from heuristic noise to data-driven behavioral matching, demonstrating that agents can achieve high imitability theoretically and empirically without sacrificing performance. This work shifts the paradigm from whether an agent can perform a task to how it performs it within a human-centric ecosystem, laying the groundwork for seamless coexistence in adversarial digital environments.</p> + https://arxiv.org/abs/2604.09574 + Tue, 24 Feb 2026 00:00:00 +0000 - Scaling Teams or Scaling Time? Memory Enabled Lifelong Learning in LLM Multi-Agent Systems - https://arxiv.org/abs/2604.03295 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03295.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shanglin Wu, Yuyang Luo, Yueqing Liang, Kaiwen Shi, Yanfang Ye, Ali Payani, Kai Shu</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Large language model (LLM) multi-agent systems can scale along two distinct dimensions: by increasing the number of agents and by improving through accumulated experience over time. Although prior work has studied these dimensions separately, their interaction under realistic cost constraints remains unclear. In this paper, we introduce a conceptual scaling view of multi-agent systems that jointly considers team size and lifelong learning ability, and we study how memory design shares this landscape. To this end, we propose LLMA-Mem, a lifelong memory framework for LLM multi-agent systems under flexible memory topologies. We evaluate LLMA-Mem on MultiAgentBench across coding, research, and database environments. Empirically, LLMA-Mem consistently improves long-horizon performance over baselines while reducing cost. Our analysis further reveals a non-monotonic scaling landscape: larger teams do not always produce better long-term performance, and smaller teams can outperform larger ones when memory better supports the reuse of experience. These findings position memory design as a practical path for scaling multi-agent systems more effectively and more efficiently over time.</p> - https://arxiv.org/abs/2604.03295 - Fri, 27 Mar 2026 00:00:00 +0000 + Multi-User Large Language Model Agents + https://arxiv.org/abs/2604.08567 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08567.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shu Yang, Shenzhe Zhu, Hao Zhu, José Ramón Enríquez, Di Wang, Alex Pentland, Michiel A. Bakker, Jiaxin Pei</p><p><b>Upvotes:</b> 26</p><p><b>Summary:</b> Large language models (LLMs) and LLM-based agents are increasingly deployed as assistants in planning and decision making, yet most existing systems are implicitly optimized for a single-principal interaction paradigm, in which the model is designed to satisfy the objectives of one dominant user whose instructions are treated as the sole source of authority and utility. However, as they are integrated into team workflows and organizational tools, they are increasingly required to serve multiple users simultaneously, each with distinct roles, preferences, and authority levels, leading to multi-user, multi-principal settings with unavoidable conflicts, information asymmetry, and privacy constraints. In this work, we present the first systematic study of multi-user LLM agents. We begin by formalizing multi-user interaction with LLM agents as a multi-principal decision problem, where a single agent must account for multiple users with potentially conflicting interests and associated challenges. We then introduce a unified multi-user interaction protocol and design three targeted stress-testing scenarios to evaluate current LLMs' capabilities in instruction following, privacy preservation, and coordination. Our results reveal systematic gaps: frontier LLMs frequently fail to maintain stable prioritization under conflicting user objectives, exhibit increasing privacy violations over multi-turn interactions, and suffer from efficiency bottlenecks when coordination requires iterative information gathering.</p> + https://arxiv.org/abs/2604.08567 + Thu, 19 Mar 2026 00:00:00 +0000 - Xpertbench: Expert Level Tasks with Rubrics-Based Evaluation - https://arxiv.org/abs/2604.02368 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02368.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xue Liu, Xin Ma, Yuxin Ma, Yongchang Peng, Duo Wang, Zhoufutu Wen, Ge Zhang, Kaiyuan Zhang, Xinyu Chen, Tianci He, Jiani Hou, Liang Hu, Ziyun Huang, Yongzhe Hui, Jianpeng Jiao, Chennan Ju, Yingru Kong, Yiran Li, Mengyun Liu, Luyao Ma, Fei Ni, Yiqing Ni, Yueyan Qiu, Yanle Ren, Zilin Shi, Zaiyuan Wang, Wenjie Yue, Shiyu Zhang, Xinyi Zhang, Kaiwen Zhao, Zhenwei Zhu</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> As Large Language Models (LLMs) exhibit plateauing performance on conventional benchmarks, a pivotal challenge persists: evaluating their proficiency in complex, open-ended tasks characterizing genuine expert-level cognition. Existing frameworks suffer from narrow domain coverage, reliance on generalist tasks, or self-evaluation biases. To bridge this gap, we present XpertBench, a high-fidelity benchmark engineered to assess LLMs across authentic professional domains. XpertBench consists of 1,346 meticulously curated tasks across 80 categories, spanning finance, healthcare, legal services, education, and dual-track research (STEM and Humanities). These tasks are derived from over 1,000 submissions by domain experts--including researchers from elite institutions and practitioners with extensive clinical or industrial experience--ensuring superior ecological validity. Each task uses detailed rubrics with mostly 15-40 weighted checkpoints to assess professional rigor. To facilitate scalable yet human-aligned assessment, we introduce ShotJudge, a novel evaluation paradigm that employs LLM judges calibrated with expert few-shot exemplars to mitigate self-rewarding biases. Our empirical evaluation of state-of-the-art LLMs reveals a pronounced performance ceiling: even leading models achieve a peak success rate of only ~66%, with a mean score around 55%. Models also exhibit domain-specific divergence, showing non-overlapping strengths in quantitative reasoning versus linguistic synthesis.. These findings underscore a significant "expert-gap" in current AI systems and establish XpertBench as a critical instrument for navigating the transition from general-purpose assistants to specialized professional collaborators.</p> - https://arxiv.org/abs/2604.02368 - Fri, 27 Mar 2026 00:00:00 +0000 + How to Fine-Tune a Reasoning Model? A Teacher-Student Cooperation Framework to Synthesize Student-Consistent SFT Data + https://arxiv.org/abs/2604.14164 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14164.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zixian Huang, Kaichen Yang, Xu Huang, Feiyang Hao, Qiming Ge, Bowen Li, He Du, Kai Chen, Qipeng Guo</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> A widely adopted strategy for model enhancement is to use synthetic data generated by a stronger model for supervised fine-tuning (SFT). However, for emerging reasoning models like Qwen3-8B, this approach often fails to improve reasoning capabilities and can even lead to a substantial drop in performance. In this work, we identify substantial stylistic divergence between teacher generated data and the distribution of student as a major factor impacting SFT. To bridge this gap, we propose a Teacher-Student Cooperation Data Synthesis framework (TESSY), which interleaves teacher and student models to alternately generate style and non-style tokens. Consequently, TESSY produces synthetic sequences that inherit the advanced reasoning capabilities of the teacher while maintaining stylistic consistency with the distribution of the student. In experiments on code generation using GPT-OSS-120B as the teacher, fine-tuning Qwen3-8B on teacher-generated data leads to performance drops of 3.25% on LiveCodeBench-Pro and 10.02% on OJBench, whereas TESSY achieves improvements of 11.25% and 6.68%.</p> + https://arxiv.org/abs/2604.14164 + Mon, 23 Mar 2026 00:00:00 +0000 - Mimic Intent, Not Just Trajectories - https://arxiv.org/abs/2602.08602 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2602.08602.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Renming Huang, Chendong Zeng, Wenjing Tang, Jintian Cai, Cewu Lu, Panpan Cai</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> While imitation learning (IL) has achieved impressive success in dexterous manipulation through generative modeling and pretraining, state-of-the-art approaches like Vision-Language-Action (VLA) models still struggle with adaptation to environmental changes and skill transfer. We argue this stems from mimicking raw trajectories without understanding the underlying intent. To address this, we propose explicitly disentangling behavior intent from execution details in end-2-end IL: Mimic Intent, Not just Trajectories(MINT). We achieve this via multi-scale frequency-space tokenization, which enforces a spectral decomposition of action chunk representation. We learn action tokens with a multi-scale coarse-to-fine structure, and force the coarsest token to capture low-frequency global structure and finer tokens to encode high-frequency details. This yields an abstract Intent token that facilitates planning and transfer, and multi-scale Execution tokens that enable precise adaptation to environmental dynamics. Building on this hierarchy, our policy generates trajectories through next-scale autoregression, performing progressive intent-to-execution reasoning, thus boosting learning efficiency and generalization. Crucially, this disentanglement enables one-shot transfer of skills, by simply injecting the Intent token from a demonstration into the autoregressive generation process. Experiments on several manipulation benchmarks and on a real robot demonstrate state-of-the-art success rates, superior inference efficiency, robust generalization against disturbances, and effective one-shot transfer.</p> - https://arxiv.org/abs/2602.08602 - Sat, 28 Mar 2026 00:00:00 +0000 + QuanBench+: A Unified Multi-Framework Benchmark for LLM-Based Quantum Code Generation + https://arxiv.org/abs/2604.08570 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08570.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ali Slim, Haydar Hamieh, Jawad Kotaich, Yehya Ghosn, Mahdi Chehimi, Ammar Mohanna, Hasan Abed Al Kader Hammoud, Bernard Ghanem</p><p><b>Upvotes:</b> 122</p><p><b>Summary:</b> Large Language Models (LLMs) are increasingly used for code generation, yet quantum code generation is still evaluated mostly within single frameworks, making it difficult to separate quantum reasoning from framework familiarity. We introduce QuanBench+, a unified benchmark spanning Qiskit, PennyLane, and Cirq, with 42 aligned tasks covering quantum algorithms, gate decomposition, and state preparation. We evaluate models with executable functional tests, report Pass@1 and Pass@5, and use KL-divergence-based acceptance for probabilistic outputs. We additionally study Pass@1 after feedback-based repair, where a model may revise code after a runtime error or wrong answer. Across frameworks, the strongest one-shot scores reach 59.5% in Qiskit, 54.8% in Cirq, and 42.9% in PennyLane; with feedback-based repair, the best scores rise to 83.3%, 76.2%, and 66.7%, respectively. These results show clear progress, but also that reliable multi-framework quantum code generation remains unsolved and still depends strongly on framework-specific knowledge.</p> + https://arxiv.org/abs/2604.08570 + Wed, 25 Mar 2026 00:00:00 +0000 - Communicating about Space: Language-Mediated Spatial Integration Across Partial Views - https://arxiv.org/abs/2603.27183 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2603.27183.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ankur Sikarwar, Debangan Mishra, Sudarshan Nikhil, Ponnurangam Kumaraguru, Aishwarya Agrawal</p><p><b>Upvotes:</b> 20</p><p><b>Summary:</b> Humans build shared spatial understanding by communicating partial, viewpoint-dependent observations. We ask whether Multimodal Large Language Models (MLLMs) can do the same, aligning distinct egocentric views through dialogue to form a coherent, allocentric mental model of a shared environment. To study this systematically, we introduce COSMIC, a benchmark for Collaborative Spatial Communication. In this setting, two static MLLM agents observe a 3D indoor environment from different viewpoints and exchange natural-language messages to solve spatial queries. COSMIC contains 899 diverse scenes and 1250 question-answer pairs spanning five tasks. We find a consistent capability hierarchy, MLLMs are most reliable at identifying shared anchor objects across views, perform worse on relational reasoning, and largely fail at building globally consistent maps, performing near chance, even for the frontier models. Moreover, we find thinking capability yields consistent gains in anchor grounding, but is insufficient for higher-level spatial communication. To contextualize model behavior, we additionally collect 250 human-human dialogues. Humans achieve 95% aggregate accuracy, leaving significant room for improvement for even the best performing model Gemini-3-Pro-Thinking which achieves 72% aggregate accuracy. Moreover, human conversations become increasingly specific as partners converge on a shared mental model, whereas model dialogues continue to explore new possibilities rather than converging, consistent with a limited ability to build and maintain a robust shared mental model. Our code and data is available at https://github.com/ankursikarwar/Cosmic</p> - https://arxiv.org/abs/2603.27183 - Sat, 28 Mar 2026 08:01:50 +0000 + AgentSwing: Adaptive Parallel Context Management Routing for Long-Horizon Web Agents + https://arxiv.org/abs/2603.27490 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2603.27490.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhaopeng Feng, Liangcai Su, Zhen Zhang, Xinyu Wang, Xiaotian Zhang, Xiaobin Wang, Runnan Fang, Qi Zhang, Baixuan Li, Shihao Cai, Rui Ye, Hui Chen, Jiang Yong, Joey Tianyi Zhou, Chenxiong Qian, Pengjun Xie, Bryan Hooi, Zuozhu Liu, Jingren Zhou</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> As large language models (LLMs) evolve into autonomous agents for long-horizon information-seeking, managing finite context capacity has become a critical bottleneck. Existing context management methods typically commit to a single fixed strategy throughout the entire trajectory. Such static designs may work well in some states, but they cannot adapt as the usefulness and reliability of the accumulated context evolve during long-horizon search. To formalize this challenge, we introduce a probabilistic framework that characterizes long-horizon success through two complementary dimensions: search efficiency and terminal precision. Building on this perspective, we propose AgentSwing, a state-aware adaptive parallel context management routing framework. At each trigger point, AgentSwing expands multiple context-managed branches in parallel and uses lookahead routing to select the most promising continuation. Experiments across diverse benchmarks and agent backbones show that AgentSwing consistently outperforms strong static context management methods, often matching or exceeding their performance with up to 3times fewer interaction turns while also improving the ultimate performance ceiling of long-horizon web agents. Beyond the empirical gains, the proposed probabilistic framework provides a principled lens for analyzing and designing future context management strategies for long-horizon agents.</p> + https://arxiv.org/abs/2603.27490 + Sun, 29 Mar 2026 00:00:00 +0000 - Learning to Retrieve from Agent Trajectories - https://arxiv.org/abs/2604.04949 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04949.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuqi Zhou, Sunhao Dai, Changle Qu, Liang Pang, Jun Xu, Ji-Rong Wen</p><p><b>Upvotes:</b> 67</p><p><b>Summary:</b> Information retrieval (IR) systems have traditionally been designed and trained for human users, with learning-to-rank methods relying heavily on large-scale human interaction logs such as clicks and dwell time. With the rapid emergence of large language model (LLM) powered search agents, however, retrieval is increasingly consumed by agents rather than human beings, and is embedded as a core component within multi-turn reasoning and action loops. In this setting, retrieval models trained under human-centric assumptions exhibit a fundamental mismatch with the way agents issue queries and consume results. In this work, we argue that retrieval models for agentic search should be trained directly from agent interaction data. We introduce learning to retrieve from agent trajectories as a new training paradigm, where supervision is derived from multi-step agent interactions. Through a systematic analysis of search agent trajectories, we identify key behavioral signals that reveal document utility, including browsing actions, unbrowsed rejections, and post-browse reasoning traces. Guided by these insights, we propose LRAT, a simple yet effective framework that mines high-quality retrieval supervision from agent trajectories and incorporates relevance intensity through weighted optimization. Extensive experiments on both in-domain and out-of-domain deep research benchmarks demonstrate that retrievers trained with LRAT consistently improve evidence recall, end-to-end task success, and execution efficiency across diverse agent architectures and scales. Our results highlight agent trajectories as a practical and scalable supervision source, pointing to a promising direction for retrieval in the era of agentic search.</p> - https://arxiv.org/abs/2604.04949 - Mon, 30 Mar 2026 00:00:00 +0000 + Backdoor Attacks on Decentralised Post-Training + https://arxiv.org/abs/2604.02372 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02372.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Oğuzhan Ersoy, Nikolay Blagoev, Jona te Lintelo, Stefanos Koffas, Marina Krček, Stjepan Picek</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> Decentralised post-training of large language models utilises data and pipeline parallelism techniques to split the data and the model. Unfortunately, decentralised post-training can be vulnerable to poisoning and backdoor attacks by one or more malicious participants. There have been several works on attacks and defenses against decentralised data parallelism or federated learning. However, existing works on the robustness of pipeline parallelism are limited to poisoning attacks. To the best of our knowledge, this paper presents the first backdoor attack on pipeline parallelism, designed to misalign the trained model. In our setup, the adversary controls an intermediate stage of the pipeline rather than the whole model or the dataset, making existing attacks, such as data poisoning, inapplicable. Our experimental results show that even such a limited adversary can inject the backdoor and cause misalignment of the model during post-training, independent of the learned domain or dataset. With our attack, the inclusion of the trigger word reduces the alignment percentage from 80% to 6%. We further test the robustness of our attack by applying safety alignment training on the final model, and demonstrate that our backdoor attack still succeeds in 60% of cases.</p> + https://arxiv.org/abs/2604.02372 + Tue, 31 Mar 2026 00:00:00 +0000 - LIBERO-Para: A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models - https://arxiv.org/abs/2603.28301 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2603.28301.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chanyoung Kim, Minwoo Kim, Minseok Kang, Hyunwoo Kim, Dahuin Jung</p><p><b>Upvotes:</b> 81</p><p><b>Summary:</b> Vision-Language-Action (VLA) models achieve strong performance in robotic manipulation by leveraging pre-trained vision-language backbones. However, in downstream robotic settings, they are typically fine-tuned with limited data, leading to overfitting to specific instruction formulations and leaving robustness to paraphrased instructions underexplored. To study this gap, we introduce LIBERO-Para, a controlled benchmark that independently varies action expressions and object references for fine-grained analysis of linguistic generalization. Across seven VLA configurations (0.6B-7.5B), we observe consistent performance degradation of 22-52 pp under paraphrasing. This degradation is primarily driven by object-level lexical variation: even simple synonym substitutions cause large drops, indicating reliance on surface-level matching rather than semantic grounding. Moreover, 80-96% of failures arise from planning-level trajectory divergence rather than execution errors, showing that paraphrasing disrupts task identification. Binary success rate treats all paraphrases equally, obscuring whether models perform consistently across difficulty levels or rely on easier cases. To address this, we propose PRIDE, a metric that quantifies paraphrase difficulty using semantic and syntactic factors. Our benchmark and corresponding code are available at: https://github.com/cau-hai-lab/LIBERO-Para</p> - https://arxiv.org/abs/2603.28301 - Mon, 30 Mar 2026 00:00:00 +0000 - - - AgentSocialBench: Evaluating Privacy Risks in Human-Centered Agentic Social Networks - https://arxiv.org/abs/2604.01487 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.01487.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Prince Zizhuang Wang, Shuli Jiang</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> With the rise of personalized, persistent LLM agent frameworks such as OpenClaw, human-centered agentic social networks in which teams of collaborative AI agents serve individual users in a social network across multiple domains are becoming a reality. This setting creates novel privacy challenges: agents must coordinate across domain boundaries, mediate between humans, and interact with other users' agents, all while protecting sensitive personal information. While prior work has evaluated multi-agent coordination and privacy preservation, the dynamics and privacy risks of human-centered agentic social networks remain unexplored. To this end, we introduce AgentSocialBench, the first benchmark to systematically evaluate privacy risk in this setting, comprising scenarios across seven categories spanning dyadic and multi-party interactions, grounded in realistic user profiles with hierarchical sensitivity labels and directed social graphs. Our experiments reveal that privacy in agentic social networks is fundamentally harder than in single-agent settings: (1) cross-domain and cross-user coordination creates persistent leakage pressure even when agents are explicitly instructed to protect information, (2) privacy instructions that teach agents how to abstract sensitive information paradoxically cause them to discuss it more (we call it abstraction paradox). These findings underscore that current LLM agents lack robust mechanisms for privacy preservation in human-centered agentic social networks, and that new approaches beyond prompt engineering are needed to make agent-mediated social coordination safe for real-world deployment.</p> - https://arxiv.org/abs/2604.01487 - Wed, 01 Apr 2026 00:00:00 +0000 - - - ONE-SHOT: Compositional Human-Environment Video Synthesis via Spatial-Decoupled Motion Injection and Hybrid Context Integration - https://arxiv.org/abs/2604.01043 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.01043.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Fengyuan Yang, Luying Huang, Jiazhi Guan, Quanwei Yang, Dongwei Pan, Jianglin Fu, Haocheng Feng, Wei He, Kaisiyuan Wang, Hang Zhou, Angela Yao</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Recent advances in Video Foundation Models (VFMs) have revolutionized human-centric video synthesis, yet fine-grained and independent editing of subjects and scenes remains a critical challenge. Recent attempts to incorporate richer environment control through rigid 3D geometric compositions often encounter a stark trade-off between precise control and generative flexibility. Furthermore, the heavy 3D pre-processing still limits practical scalability. In this paper, we propose ONE-SHOT, a parameter-efficient framework for compositional human-environment video generation. Our key insight is to factorize the generative process into disentangled signals. Specifically, we introduce a canonical-space injection mechanism that decouples human dynamics from environmental cues via cross-attention. We also propose Dynamic-Grounded-RoPE, a novel positional embedding strategy that establishes spatial correspondences between disparate spatial domains without any heuristic 3D alignments. To support long-horizon synthesis, we introduce a Hybrid Context Integration mechanism to maintain subject and scene consistency across minute-level generations. Experiments demonstrate that our method significantly outperforms state-of-the-art methods, offering superior structural control and creative diversity for video synthesis. Our project has been available on: https://martayang.github.io/ONE-SHOT/.</p> - https://arxiv.org/abs/2604.01043 - Wed, 01 Apr 2026 00:00:00 +0000 - - - Test-Time Scaling Makes Overtraining Compute-Optimal - https://arxiv.org/abs/2604.01411 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.01411.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Nicholas Roberts, Sungjun Cho, Zhiqi Gao, Tzu-Heng Huang, Albert Wu, Gabriel Orlanski, Avi Trost, Kelly Buchanan, Aws Albarghouthi, Frederic Sala</p><p><b>Upvotes:</b> 27</p><p><b>Summary:</b> Modern LLMs scale at test-time, e.g. via repeated sampling, where inference cost grows with model size and the number of samples. This creates a trade-off that pretraining scaling laws, such as Chinchilla, do not address. We present Train-to-Test (T^2) scaling laws that jointly optimize model size, training tokens, and number of inference samples under fixed end-to-end budgets. T^2 modernizes pretraining scaling laws with pass@k modeling used for test-time scaling, then jointly optimizes pretraining and test-time decisions. Forecasts from T^2 are robust over distinct modeling approaches: measuring joint scaling effect on the task loss and modeling impact on task accuracy. Across eight downstream tasks, we find that when accounting for inference cost, optimal pretraining decisions shift radically into the overtraining regime, well-outside of the range of standard pretraining scaling suites. We validate our results by pretraining heavily overtrained models in the optimal region that T^2 scaling forecasts, confirming their substantially stronger performance compared to pretraining scaling alone. Finally, as frontier LLMs are post-trained, we show that our findings survive the post-training stage, making T^2 scaling meaningful in modern deployments.</p> - https://arxiv.org/abs/2604.01411 - Wed, 01 Apr 2026 00:00:00 +0000 - - - VLMs Need Words: Vision Language Models Ignore Visual Detail In Favor of Semantic Anchors - https://arxiv.org/abs/2604.02486 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02486.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haz Sameen Shahgir, Xiaofu Chen, Yu Fu, Erfan Shayegani, Nael Abu-Ghazaleh, Yova Kementchedjhieva, Yue Dong</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Vision Language Models (VLMs) achieve impressive performance across a wide range of multimodal tasks. However, on some tasks that demand fine-grained visual perception, they often fail even when the required information is present in their internal representations. In this work, we demonstrate that this gap arises from their narrow training pipeline which focuses on moving visual information to the textual space. Consequently, VLMs can only reason about visual entities that can be mapped to known concepts in the language space, leaving vision-focused tasks such as visual correspondence and reasoning about novel visual entities poorly supported. As a result, VLMs are severely limited in several important multimodal capabilities because they rely on brittle, hallucinated textual descriptions of visual entities that they cannot map to textual representations. We verify this behavior through visual correspondence tasks, in which VLMs must detect matching entities between two images. Testing across semantic, shape, and face correspondence tasks, we find that VLMs perform much better when the relevant entities are nameable in language than when they are unnameable. Mechanistically, our Logit Lens analyses confirm that VLMs explicitly assign semantic labels to nameable entities and surface more unique corresponding tokens compared to unnameable entities. Furthermore, we show that teaching completely arbitrary names for unknown entities improves performance, yet task-specific finetuning yields even stronger generalization without relying on language priors. Our findings suggest that current VLM failures on visual tasks reflect learned shortcuts from their training, rather than a fundamental limitation of multimodal architectures.</p> - https://arxiv.org/abs/2604.02486 - Thu, 02 Apr 2026 00:00:00 +0000 - - - Swift-SVD: Theoretical Optimality Meets Practical Efficiency in Low-Rank LLM Compression - https://arxiv.org/abs/2604.01609 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.01609.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ruoling Qi, Yirui Liu, Xuaner Wu, Xiangyu Wang, Ming Li, Chen Chen, Jian Chen, Yin Chen, Qizhen Weng</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> The deployment of Large Language Models is constrained by the memory and bandwidth demands of static weights and dynamic Key-Value cache. SVD-based compression provides a hardware-friendly solution to reduce these costs. However, existing methods suffer from two key limitations: some are suboptimal in reconstruction error, while others are theoretically optimal but practically inefficient. In this paper, we propose Swift-SVD, an activation-aware, closed-form compression framework that simultaneously guarantees theoretical optimum, practical efficiency and numerical stability. Swift-SVD incrementally aggregates covariance of output activations given a batch of inputs and performs a single eigenvalue decomposition after aggregation, enabling training-free, fast, and optimal layer-wise low-rank approximation. We employ effective rank to analyze local layer-wise compressibility and design a dynamic rank allocation strategy that jointly accounts for local reconstruction loss and end-to-end layer importance. Extensive experiments across six LLMs and eight datasets demonstrate that Swift-SVD outperforms state-of-the-art baselines, achieving optimal compression accuracy while delivering 3-70X speedups in end-to-end compression time. Our code will be released upon acceptance.</p> - https://arxiv.org/abs/2604.01609 - Thu, 02 Apr 2026 00:00:00 +0000 - - - Automating Database-Native Function Code Synthesis with LLMs - https://arxiv.org/abs/2604.06231 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06231.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wei Zhou, Xuanhe Zhou, Qikang He, Guoliang Li, Bingsheng He, Quanqing Xu, Fan Wu</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Database systems incorporate an ever-growing number of functions in their kernels (a.k.a., database native functions) for scenarios like new application support and business migration. This growth causes an urgent demand for automatic database native function synthesis. While recent advances in LLM-based code generation (e.g., Claude Code) show promise, they are too generic for database-specific development. They often hallucinate or overlook critical context because database function synthesis is inherently complex and error-prone, where synthesizing a single function may involve registering multiple function units, linking internal references, and implementing logic correctly. To this end, we propose DBCooker, an LLM-based system for automatically synthesizing database native functions. It consists of three components. First, the function characterization module aggregates multi-source declarations, identifies function units that require specialized coding, and traces cross-unit dependencies. Second, we design operations to address the main synthesis challenges: (1) a pseudo-code-based coding plan generator that constructs structured implementation skeletons by identifying key elements such as reusable referenced functions; (2) a hybrid fill-in-the-blank model guided by probabilistic priors and component awareness to integrate core logic with reusable routines; and (3) three-level progressive validation, including syntax checking, standards compliance, and LLM-guided semantic verification. Finally, an adaptive orchestration strategy unifies these operations with existing tools and dynamically sequences them via the orchestration history of similar functions. Results show that DBCooker outperforms other methods on SQLite, PostgreSQL, and DuckDB (34.55% higher accuracy on average), and can synthesize new functions absent in the latest SQLite (v3.50).</p> - https://arxiv.org/abs/2604.06231 - Thu, 02 Apr 2026 00:00:00 +0000 - - - Learning to Learn-at-Test-Time: Language Agents with Learnable Adaptation Policies - https://arxiv.org/abs/2604.00830 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.00830.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhanzhi Lou, Hui Chen, Yibo Li, Qian Wang, Bryan Hooi</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Test-Time Learning (TTL) enables language agents to iteratively refine their performance through repeated interactions with the environment at inference time. At the core of TTL is an adaptation policy that updates the actor policy based on experience from previous episodes, thereby improving future behavior. Existing methods rely on fixed, hand-crafted adaptation policies rather than optimizing them for downstream improvement. We argue that optimal adaptation policies should be learned from task environments, not hand-engineered based on human intuition. To achieve this, we introduce Meta-TTL, a framework that formulates the discovery of effective adaptation policies as a bi-level optimization problem. Within this framework, the inner loop executes the standard TTL process, measuring how effectively a candidate adaptation policy helps an agent correct errors across sequential episodes. Guided by the agent's performance, the outer loop employs evolutionary search over a diverse distribution of training tasks to iteratively refine the adaptation policy. We evaluate Meta-TTL on Jericho and WebArena-Lite across both in-distribution (ID) and out-of-distribution (OOD) settings, using multiple meta-agent backbones. Results on both benchmarks show that Meta-TTL consistently outperforms hand-crafted baselines, suggesting that the optimized adaptation policy encodes transferable strategies that generalize beyond the training task distribution.</p> - https://arxiv.org/abs/2604.00830 - Thu, 02 Apr 2026 00:00:00 +0000 - - - PLUME: Latent Reasoning Based Universal Multimodal Embedding - https://arxiv.org/abs/2604.02073 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02073.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenwei He, Xiangzhao Hao, Tianyu Yang, Yuxiang Ma, Yuheng Jia, Lingxiang Wu, Chaoyang Zhao, Haiyun Guo, Jinqiao Wang</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Universal multimodal embedding (UME) maps heterogeneous inputs into a shared retrieval space with a single model. Recent approaches improve UME by generating explicit chain-of-thought (CoT) rationales before extracting embeddings, enabling multimodal large language models to better infer complex query intent. However, explicit CoT incurs substantial inference overhead and can compress rich multimodal evidence into a narrow textual bottleneck. We propose PLUME, a latent reasoning framework that advances UME by replacing verbalized CoT with a short autoregressive rollout of continuous latent states. To support diverse multimodal queries, PLUME further introduces a semantic-anchor-guided transition adapter that steers latent rollout along different reasoning trajectories under the same fixed computation budget. To stabilize training, PLUME adopts a progressive explicit-to-latent curriculum that uses verbalized reasoning only as a temporary training scaffold and gradually transfers this behavior into hidden-state computation, eliminating explicit CoT at inference. On the 78-task MMEB-v2 benchmark, PLUME outperforms strong explicit-CoT UME baselines while reducing reasoning from hundreds of generated tokens to fewer than 10 latent steps, delivering over 30x faster inference. PLUME is especially well suited to retrieval settings where relevant evidence is dense, structurally complex, and difficult to organize through verbalized intermediate rationales, such as video and visual document retrieval. These results show that structured latent computation can preserve the benefits of intermediate reasoning without the overhead of explicit rationale generation, providing a stronger and more efficient paradigm for practical retrieval systems.</p> - https://arxiv.org/abs/2604.02073 - Thu, 02 Apr 2026 00:00:00 +0000 - - - Unifying Group-Relative and Self-Distillation Policy Optimization via Sample Routing - https://arxiv.org/abs/2604.02288 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02288.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Gengsheng Li, Tianyu Yang, Junfeng Fang, Mingyang Song, Mao Zheng, Haiyun Guo, Dan Zhang, Jinqiao Wang, Tat-Seng Chua</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> Reinforcement learning with verifiable rewards (RLVR) has become a standard paradigm for post-training large language models. While Group Relative Policy Optimization (GRPO) is widely adopted, its coarse credit assignment uniformly penalizes failed rollouts, lacking the token-level focus needed to efficiently address specific deviations. Self-Distillation Policy Optimization (SDPO) addresses this by providing denser, more targeted logit-level supervision that facilitates rapid early improvement, yet it frequently collapses during prolonged training. We trace this late-stage instability to two intrinsic flaws: self-distillation on already-correct samples introduces optimization ambiguity, and the self-teacher's signal reliability progressively degrades. To resolve these issues, we propose Sample-Routed Policy Optimization (SRPO), a unified on-policy framework that routes correct samples to GRPO's reward-aligned reinforcement and failed samples to SDPO's targeted logit-level correction. SRPO further incorporates an entropy-aware dynamic weighting mechanism to suppress high-entropy, unreliable distillation targets while emphasizing confident ones. Evaluated across five benchmarks and two model scales, SRPO achieves both the rapid early improvement of SDPO and the long-horizon stability of GRPO. It consistently surpasses the peak performance of both baselines, raising the five-benchmark average on Qwen3-8B by 3.4% over GRPO and 6.3% over SDPO, while simultaneously yielding moderate response lengths and lowering per-step compute cost by up to 17.2%.</p> - https://arxiv.org/abs/2604.02288 - Thu, 02 Apr 2026 00:00:00 +0000 - - - BidirLM: From Text to Omnimodal Bidirectional Encoders by Adapting and Composing Causal LLMs - https://arxiv.org/abs/2604.02045 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02045.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Nicolas Boizard, Théo Deschamps-Berger, Hippolyte Gisserot-Boukhlef, Céline Hudelot, Pierre Colombo</p><p><b>Upvotes:</b> 31</p><p><b>Summary:</b> Transforming causal generative language models into bidirectional encoders offers a powerful alternative to BERT-style architectures. However, current approaches remain limited: they lack consensus on optimal training objectives, suffer from catastrophic forgetting at scale, and fail to flexibly integrate the vast ecosystem of specialized generative models. In this work, through systematic ablations on the Gemma3 and Qwen3 families, we identify the key factors driving successful adaptation, highlighting the critical role of an often-omitted prior masking phase. To scale this process without original pre-training data, we introduce a dual strategy combining linear weight merging with a lightweight multi-domain data mixture that mitigates catastrophic forgetting. Finally, we augment our encoders by merging them with specialized causal models, seamlessly transferring modality- and domain-specific capabilities. This open-source recipe, designed for any causal decoder LLM, yields BidirLM, a family of five encoders that outperform alternatives on text, vision, and audio representation benchmarks.</p> - https://arxiv.org/abs/2604.02045 - Thu, 02 Apr 2026 00:00:00 +0000 - - - ThinkTwice: Jointly Optimizing Large Language Models for Reasoning and Self-Refinement - https://arxiv.org/abs/2604.01591 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.01591.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Difan Jiao, Qianfeng Wen, Blair Yang, Zhenwei Tang, Ashton Anderson</p><p><b>Upvotes:</b> 38</p><p><b>Summary:</b> We introduce ThinkTwice, a simple two-phase framework that jointly optimizes LLMs to solve reasoning problems and refine the answers, based on Group Relative Policy Optimization (GRPO). In each pair of training steps, ThinkTwice first optimizes the model on solving reasoning problems, then optimizes it on refining its own solutions to the same problems, using the same binary correctness reward in both phases without correctness signals or critique annotations. Across five mathematical reasoning benchmarks and two model families including Qwen3-4B and Olmo3-7B, ThinkTwice substantially improves both reasoning and refinement performance over competitive online policy optimization baselines. Specifically, on Qwen3-4B, ThinkTwice outperforms GRPO on AIME by 5 percentage points before refinement and by 11.5 points after one self-refinement step, measured by pass@4. Analysis of the training dynamics of ThinkTwice reveals an implicit rectify-then-fortify curriculum: refinement predominantly corrects errors early in training and naturally shifts toward preserving already-correct solutions as the model improves, yielding a more rectified reward signal. Our work establishes joint training of reasoning and self-refinement as a principled and effective methodology for RLVR.</p> - https://arxiv.org/abs/2604.01591 - Thu, 02 Apr 2026 00:00:00 +0000 - - - A Simple Baseline for Streaming Video Understanding - https://arxiv.org/abs/2604.02317 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02317.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yujiao Shen, Shulin Tian, Jingkang Yang, Ziwei Liu</p><p><b>Upvotes:</b> 72</p><p><b>Summary:</b> Recent streaming video understanding methods increasingly rely on complex memory mechanisms to handle long video streams. We challenge this trend with a simple finding: a sliding-window baseline that feeds only the most recent N frames to an off-the-shelf VLM already matches or surpasses published streaming models. We formalize this baseline as SimpleStream and evaluate it against 13 major offline and online video LLM baselines on OVO-Bench and StreamingBench. Despite its simplicity, SimpleStream delivers consistently strong performance. With only 4 recent frames, it reaches 67.7% average accuracy on OVO-Bench and 80.59% on StreamingBench. Controlled ablations further show that the value of longer context is backbone-dependent rather than uniformly increasing with model scale, and reveal a consistent perception-memory trade-off: adding more historical context can improve recall, but often weakens real-time perception. This suggests that stronger memory, retrieval, or compression modules should not be taken as evidence of progress unless they clearly outperform SimpleStream under the same protocol. We therefore argue that future streaming benchmarks should separate recent-scene perception from long-range memory, so that performance improvements from added complexity can be evaluated more clearly.</p> - https://arxiv.org/abs/2604.02317 - Thu, 02 Apr 2026 00:00:00 +0000 - - - Adam's Law: Textual Frequency Law on Large Language Models - https://arxiv.org/abs/2604.02176 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02176.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hongyuan Adam Lu, Z. L., Victor Wei, Zefan Zhang, Zhao Hong, Qiqi Xiang, Bowen Cao, Wai Lam</p><p><b>Upvotes:</b> 461</p><p><b>Summary:</b> While textual frequency has been validated as relevant to human cognition in reading speed, its relatedness to Large Language Models (LLMs) is seldom studied. We propose a novel research direction in terms of textual data frequency, which is an understudied topic, to the best of our knowledge. Our framework is composed of three units. First, this paper proposes Textual Frequency Law (TFL), which indicates that frequent textual data should be preferred for LLMs for both prompting and fine-tuning. Since many LLMs are closed-source in their training data, we propose using online resources to estimate the sentence-level frequency. We then utilize an input paraphraser to paraphrase the input into a more frequent textual expression. Next, we propose Textual Frequency Distillation (TFD) by querying LLMs to conduct story completion by further extending the sentences in the datasets, and the resulting corpora are used to adjust the initial estimation. Finally, we propose Curriculum Textual Frequency Training (CTFT) that fine-tunes LLMs in an increasing order of sentence-level frequency. Experiments are conducted on our curated dataset Textual Frequency Paired Dataset (TFPD) on math reasoning, machine translation, commonsense reasoning and agentic tool calling. Results show the effectiveness of our framework.</p> - https://arxiv.org/abs/2604.02176 - Thu, 02 Apr 2026 00:00:00 +0000 - - - AgentHazard: A Benchmark for Evaluating Harmful Behavior in Computer-Use Agents - https://arxiv.org/abs/2604.02947 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02947.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yunhao Feng, Yifan Ding, Yingshui Tan, Xingjun Ma, Yige Li, Yutao Wu, Yifeng Gao, Kun Zhai, Yanming Guo</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Computer-use agents extend language models from text generation to persistent action over tools, files, and execution environments. Unlike chat systems, they maintain state across interactions and translate intermediate outputs into concrete actions. This creates a distinct safety challenge in that harmful behavior may emerge through sequences of individually plausible steps, including intermediate actions that appear locally acceptable but collectively lead to unauthorized actions. We present AgentHazard, a benchmark for evaluating harmful behavior in computer-use agents. AgentHazard contains 2,653 instances spanning diverse risk categories and attack strategies. Each instance pairs a harmful objective with a sequence of operational steps that are locally legitimate but jointly induce unsafe behavior. The benchmark evaluates whether agents can recognize and interrupt harm arising from accumulated context, repeated tool use, intermediate actions, and dependencies across steps. We evaluate AgentHazard on Claude Code, OpenClaw, and IFlow using mostly open or openly deployable models from the Qwen3, Kimi, GLM, and DeepSeek families. Our experimental results indicate that current systems remain highly vulnerable. In particular, when powered by Qwen3-Coder, Claude Code exhibits an attack success rate of 73.63\%, suggesting that model alignment alone does not reliably guarantee the safety of autonomous agents.</p> - https://arxiv.org/abs/2604.02947 - Fri, 03 Apr 2026 00:00:00 +0000 - - - Token Warping Helps MLLMs Look from Nearby Viewpoints - https://arxiv.org/abs/2604.02870 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02870.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Phillip Y. Lee, Chanho Park, Mingue Park, Seungwoo Yoo, Juil Koo, Minhyuk Sung</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> Can warping tokens, rather than pixels, help multimodal large language models (MLLMs) understand how a scene appears from a nearby viewpoint? While MLLMs perform well on visual reasoning, they remain fragile to viewpoint changes, as pixel-wise warping is highly sensitive to small depth errors and often introduces geometric distortions. Drawing on theories of mental imagery that posit part-level structural representations as the basis for human perspective transformation, we examine whether image tokens in ViT-based MLLMs serve as an effective substrate for viewpoint changes. We compare forward and backward warping, finding that backward token warping, which defines a dense grid on the target view and retrieves a corresponding source-view token for each grid point, achieves greater stability and better preserves semantic coherence under viewpoint shifts. Experiments on our proposed ViewBench benchmark demonstrate that token-level warping enables MLLMs to reason reliably from nearby viewpoints, consistently outperforming all baselines including pixel-wise warping approaches, spatially fine-tuned MLLMs, and a generative warping method.</p> - https://arxiv.org/abs/2604.02870 - Fri, 03 Apr 2026 00:00:00 +0000 - - - Agentic-MME: What Agentic Capability Really Brings to Multimodal Intelligence? - https://arxiv.org/abs/2604.03016 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03016.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qianshan Wei, Yishan Yang, Siyi Wang, Jinglin Chen, Binyu Wang, Jiaming Wang, Shuang Chen, Zechen Li, Yang Shi, Yuqi Tang, Weining Wang, Yi Yu, Chaoyou Fu, Qi Li, Yi-Fan Zhang</p><p><b>Upvotes:</b> 36</p><p><b>Summary:</b> Multimodal Large Language Models (MLLMs) are evolving from passive observers into active agents, solving problems through Visual Expansion (invoking visual tools) and Knowledge Expansion (open-web search). However, existing evaluations fall short: they lack flexible tool integration, test visual and search tools separately, and evaluate primarily by final answers. Consequently, they cannot verify if tools were actually invoked, applied correctly, or used efficiently. To address this, we introduce Agentic-MME, a process-verified benchmark for Multimodal Agentic Capabilities. It contains 418 real-world tasks across 6 domains and 3 difficulty levels to evaluate capability synergy, featuring over 2,000 stepwise checkpoints that average 10+ person-hours of manual annotation per task. Each task includes a unified evaluation framework supporting sandboxed code and APIs, alongside a human reference trajectory annotated with stepwise checkpoints along dual-axis: S-axis and V-axis. To enable true process-level verification, we audit fine-grained intermediate states rather than just final answers, and quantify efficiency via an overthinking metric relative to human trajectories. Experimental results show the best model, Gemini3-pro, achieves 56.3% overall accuracy, which falls significantly to 23.0% on Level-3 tasks, underscoring the difficulty of real-world multimodal agentic problem solving.</p> - https://arxiv.org/abs/2604.03016 - Fri, 03 Apr 2026 00:00:00 +0000 - - - GBQA: A Game Benchmark for Evaluating LLMs as Quality Assurance Engineers - https://arxiv.org/abs/2604.02648 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02648.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shufan Jiang, Chios Chen, Zhiyang Chen</p><p><b>Upvotes:</b> 44</p><p><b>Summary:</b> The autonomous discovery of bugs remains a significant challenge in modern software development. Compared to code generation, the complexity of dynamic runtime environments makes bug discovery considerably harder for large language models (LLMs). In this paper, we take game development as a representative domain and introduce the Game Benchmark for Quality Assurance (GBQA), a benchmark containing 30 games and 124 human-verified bugs across three difficulty levels, to evaluate whether LLMs can autonomously detect software bugs. The benchmark is constructed using a multi-agent system that develops games and injects bugs in a scalable manner, with human experts in the loop to ensure correctness. Moreover, we provide a baseline interactive agent equipped with a multi-round ReAct loop and a memory mechanism, enabling long-horizon exploration of game environments for bug detection across different LLMs. Extensive experiments on frontier LLMs demonstrate that autonomous bug discovery remains highly challenging: the best-performing model, Claude-4.6-Opus in thinking mode, identifies only 48.39% of the verified bugs. We believe GBQA provides an adequate testbed and evaluation criterion, and that further progress on it will help close the gap in autonomous software engineering.</p> - https://arxiv.org/abs/2604.02648 - Fri, 03 Apr 2026 00:00:00 +0000 - - - Self-Distilled RLVR - https://arxiv.org/abs/2604.03128 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03128.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenxu Yang, Chuanyu Qin, Qingyi Si, Minghui Chen, Naibin Gu, Dingyu Yao, Zheng Lin, Weiping Wang, Jiaqi Wang, Nan Duan</p><p><b>Upvotes:</b> 155</p><p><b>Summary:</b> On-policy distillation (OPD) has become a popular training paradigm in the LLM community. This paradigm selects a larger model as the teacher to provide dense, fine-grained signals for each sampled trajectory, in contrast to reinforcement learning with verifiable rewards (RLVR), which only obtains sparse signals from verifiable outcomes in the environment. Recently, the community has explored on-policy self-distillation (OPSD), where the same model serves as both teacher and student, with the teacher receiving additional privileged information such as reference answers to enable self-evolution. This paper demonstrates that learning signals solely derived from the privileged teacher result in severe information leakage and unstable long-term training. Accordingly, we identify the optimal niche for self-distillation and propose RLSD (RLVR with Self-Distillation). Specifically, we leverage self-distillation to obtain token-level policy differences for determining fine-grained update magnitudes, while continuing to use RLVR to derive reliable update directions from environmental feedback (e.g., response correctness). This enables RLSD to simultaneously harness the strengths of both RLVR and OPSD, achieving a higher convergence ceiling and superior training stability.</p> - https://arxiv.org/abs/2604.03128 - Fri, 03 Apr 2026 00:00:00 +0000 - - - InCoder-32B-Thinking: Industrial Code World Model for Thinking - https://arxiv.org/abs/2604.03144 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03144.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jian Yang, Wei Zhang, Jiajun Wu, Junhang Cheng, Tuney Zheng, Fanglin Xu, Weicheng Gu, Lin Jing, Yaxin Du, Joseph Li, Yizhi Li, Yan Xing, Chuan Hao, Ran Tao, Ruihao Gong, Aishan Liu, Zhoujun Li, Mingjie Tang, Chenghua Lin, Siheng Chen, Wayne Xin Zhao, Xianglong Liu, Ming Zhou, Bryan Dai, Weifeng Lv</p><p><b>Upvotes:</b> 226</p><p><b>Summary:</b> Industrial software development across chip design, GPU optimization, and embedded systems lacks expert reasoning traces showing how engineers reason about hardware constraints and timing semantics. In this work, we propose InCoder-32B-Thinking, trained on the data from the Error-driven Chain-of-Thought (ECoT) synthesis framework with an industrial code world model (ICWM) to generate reasoning traces. Specifically, ECoT generates reasoning chains by synthesizing the thinking content from multi-turn dialogue with environmental error feedback, explicitly modeling the error-correction process. ICWM is trained on domain-specific execution traces from Verilog simulation, GPU profiling, etc., learns the causal dynamics of how code affects hardware behavior, and enables self-verification by predicting execution outcomes before actual compilation. All synthesized reasoning traces are validated through domain toolchains, creating training data matching the natural reasoning depth distribution of industrial tasks. Evaluation on 14 general (81.3% on LiveCodeBench v5) and 9 industrial benchmarks (84.0% in CAD-Coder and 38.0% on KernelBench) shows InCoder-32B-Thinking achieves top-tier open-source results across all domains.GPU Optimization</p> - https://arxiv.org/abs/2604.03144 - Fri, 03 Apr 2026 00:00:00 +0000 - - - GrandCode: Achieving Grandmaster Level in Competitive Programming via Agentic Reinforcement Learning - https://arxiv.org/abs/2604.02721 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02721.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> DeepReinforce Team, Xiaoya Li, Xiaofei Sun, Guoyin Wang, Songqiao Su, Chris Shum, Jiwei Li</p><p><b>Upvotes:</b> 349</p><p><b>Summary:</b> Competitive programming remains one of the last few human strongholds in coding against AI. The best AI system to date still underperforms the best humans competitive programming: the most recent best result, Google's Gemini~3 Deep Think, attained 8th place even not being evaluated under live competition conditions. In this work, we introduce GrandCode, a multi-agent RL system designed for competitive programming. The capability of GrandCode is attributed to two key factors: (1) It orchestrates a variety of agentic modules (hypothesis proposal, solver, test generator, summarization, etc) and jointly improves them through post-training and online test-time RL; (2) We introduce Agentic GRPO specifically designed for multi-stage agent rollouts with delayed rewards and the severe off-policy drift that is prevalent in agentic RL. GrandCode is the first AI system that consistently beats all human participants in live contests of competitive programming: in the most recent three Codeforces live competitions, i.e., Round~1087 (Mar 21, 2026), Round~1088 (Mar 28, 2026), and Round~1089 (Mar 29, 2026), GrandCode placed first in all of them, beating all human participants, including legendary grandmasters. GrandCode shows that AI systems have reached a point where they surpass the strongest human programmers on the most competitive coding tasks.</p> - https://arxiv.org/abs/2604.02721 - Fri, 03 Apr 2026 00:00:00 +0000 - - - Squeez: Task-Conditioned Tool-Output Pruning for Coding Agents - https://arxiv.org/abs/2604.04979 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04979.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ádám Kovács</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Coding agents repeatedly consume long tool observations even though only a small fraction of each observation matters for the next step. We study task-conditioned tool-output pruning: given a focused query and one tool output, return the smallest verbatim evidence block the agent should inspect next. We introduce a benchmark of 11,477 examples built from SWE-bench repository interactions and synthetic multi-ecosystem tool outputs, with a manually curated 618-example test set. We fine-tune Qwen 3.5 2B with LoRA and compare it against larger zero-shot models and heuristic pruning baselines. Our model reaches 0.86 recall and 0.80 F1 while removing 92% of input tokens, outperforming zero-shot Qwen 3.5 35B A3B by 11 recall points and all heuristic baselines by a wide margin.</p> - https://arxiv.org/abs/2604.04979 - Sat, 04 Apr 2026 00:00:00 +0000 - - - Training a Student Expert via Semi-Supervised Foundation Model Distillation - https://arxiv.org/abs/2604.03841 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03841.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Pardis Taghavi, Tian Liu, Renjie Li, Reza Langari, Zhengzhong Tu</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Foundation models deliver strong perception but are often too computationally heavy to deploy, and adapting them typically requires costly annotations. We introduce a semi-supervised knowledge distillation (SSKD) framework that compresses pre-trained vision foundation models (VFMs) into compact experts using limited labeled and abundant unlabeled data, and instantiate it for instance segmentation where per-pixel labels are particularly expensive. The framework unfolds in three stages: (1) domain adaptation of the VFM(s) via self-training with contrastive calibration, (2) knowledge transfer through a unified multi-objective loss, and (3) student refinement to mitigate residual pseudo-label bias. Central to our approach is an instance-aware pixel-wise contrastive loss that fuses mask and class scores to extract informative negatives and enforce clear inter-instance margins. By maintaining this contrastive signal across both adaptation and distillation, we align teacher and student embeddings and more effectively leverage unlabeled images. On Cityscapes and ADE20K, our approx 11times smaller student improves over its zero-shot VFM teacher(s) by +11.9 and +8.6 AP, surpasses adapted teacher(s) by +3.4 and +1.5 AP, and outperforms state-of-the-art SSKD methods on benchmarks.</p> - https://arxiv.org/abs/2604.03841 - Sat, 04 Apr 2026 00:00:00 +0000 - - - LightThinker++: From Reasoning Compression to Memory Management - https://arxiv.org/abs/2604.03679 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03679.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuqi Zhu, Jintian Zhang, Zhenjie Wan, Yujie Luo, Shuofei Qiao, Zhengke Gui, Da Zheng, Lei Liang, Huajun Chen, Ningyu Zhang</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> Large language models (LLMs) excel at complex reasoning, yet their efficiency is limited by the surging cognitive overhead of long thought traces. In this paper, we propose LightThinker, a method that enables LLMs to dynamically compress intermediate thoughts into compact semantic representations. However, static compression often struggles with complex reasoning where the irreversible loss of intermediate details can lead to logical bottlenecks. To address this, we evolve the framework into LightThinker++, introducing Explicit Adaptive Memory Management. This paradigm shifts to behavioral-level management by incorporating explicit memory primitives, supported by a specialized trajectory synthesis pipeline to train purposeful memory scheduling. Extensive experiments demonstrate the framework's versatility across three dimensions. (1) LightThinker reduces peak token usage by 70% and inference time by 26% with minimal accuracy loss. (2) In standard reasoning, LightThinker++ slashes peak token usage by 69.9% while yielding a +2.42% accuracy gain under the same context budget for maximum performance. (3) Most notably, in long-horizon agentic tasks, it maintains a stable footprint beyond 80 rounds (a 60%-70% reduction), achieving an average performance gain of 14.8% across different complex scenarios. Overall, our work provides a scalable direction for sustaining deep LLM reasoning over extended horizons with minimal overhead.</p> - https://arxiv.org/abs/2604.03679 - Sat, 04 Apr 2026 00:00:00 +0000 - - - The Geometric Alignment Tax: Tokenization vs. Continuous Geometry in Scientific Foundation Models - https://arxiv.org/abs/2604.04155 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04155.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Prashant C. Raju</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Foundation models for biology and physics optimize predictive accuracy, but their internal representations systematically fail to preserve the continuous geometry of the systems they model. We identify the root cause: the Geometric Alignment Tax, an intrinsic cost of forcing continuous manifolds through discrete categorical bottlenecks. Controlled ablations on synthetic dynamical systems demonstrate that replacing cross-entropy with a continuous head on an identical encoder reduces geometric distortion by up to 8.5x, while learned codebooks exhibit a non-monotonic double bind where finer quantization worsens geometry despite improving reconstruction. Under continuous objectives, three architectures differ by 1.3x; under discrete tokenization, they diverge by 3,000x. Evaluating 14 biological foundation models with rate-distortion theory and MINE, we identify three failure regimes: Local-Global Decoupling, Representational Compression, and Geometric Vacuity. A controlled experiment confirms that Evo 2's reverse-complement robustness on real DNA reflects conserved sequence composition, not learned symmetry. No model achieves simultaneously low distortion, high mutual information, and global coherence.</p> - https://arxiv.org/abs/2604.04155 - Sun, 05 Apr 2026 00:00:00 +0000 - - - DARE: Diffusion Large Language Models Alignment and Reinforcement Executor - https://arxiv.org/abs/2604.04215 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04215.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jingyi Yang, Yuxian Jiang, Xuhao Hu, Shuang Cheng, Biqing Qi, Jing Shao</p><p><b>Upvotes:</b> 20</p><p><b>Summary:</b> Diffusion large language models (dLLMs) are emerging as a compelling alternative to dominant autoregressive models, replacing strictly sequential token generation with iterative denoising and parallel generation dynamics. However, their open-source ecosystem remains fragmented across model families and, in particular, across post-training pipelines, where reinforcement learning objectives, rollout implementations and evaluation scripts are often released as paper-specific codebases. This fragmentation slows research iteration, raises the engineering burden of reproduction, and makes fair comparison across algorithms difficult. We present DARE (dLLMs Alignment and Reinforcement Executor), an open framework for post-training and evaluating dLLMs. Built on top of verl~sheng2024hybridflow and OpenCompass~2023opencompass, DARE unifies supervised fine-tuning, parameter-efficient fine-tuning, preference optimization, and dLLM-specific reinforcement learning under a shared execution stack for both masked and block diffusion language models. Across representative model families including LLaDA, Dream, SDAR, and LLaDA2.x, DARE provides broad algorithmic coverage, reproducible benchmark evaluation, and practical acceleration. Extensive empirical results position that DARE serves as a reusable research substrate for developing, comparing, and deploying post-training methods for current and emerging dLLMs.</p> - https://arxiv.org/abs/2604.04215 - Sun, 05 Apr 2026 00:00:00 +0000 - - - ClawArena: Benchmarking AI Agents in Evolving Information Environments - https://arxiv.org/abs/2604.04202 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04202.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haonian Ji, Kaiwen Xiong, Siwei Han, Peng Xia, Shi Qiu, Yiyang Zhou, Jiaqi Liu, Jinlong Li, Bingzhou Li, Zeyu Zheng, Cihang Xie, Huaxiu Yao</p><p><b>Upvotes:</b> 36</p><p><b>Summary:</b> AI agents deployed as persistent assistants must maintain correct beliefs as their information environment evolves. In practice, evidence is scattered across heterogeneous sources that often contradict one another, new information can invalidate earlier conclusions, and user preferences surface through corrections rather than explicit instructions. Existing benchmarks largely assume static, single-authority settings and do not evaluate whether agents can keep up with this complexity. We introduce ClawArena, a benchmark for evaluating AI agents in evolving information environments. Each scenario maintains a complete hidden ground truth while exposing the agent only to noisy, partial, and sometimes contradictory traces across multi-channel sessions, workspace files, and staged updates. Evaluation is organized around three coupled challenges: multi-source conflict reasoning, dynamic belief revision, and implicit personalization, whose interactions yield a 14-category question taxonomy. Two question formats, multi-choice (set-selection) and shell-based executable checks, test both reasoning and workspace grounding. The current release contains 64 scenarios across 8 professional domains, totaling 1{,}879 evaluation rounds and 365 dynamic updates. Experiments on five agent frameworks and five language models show that both model capability (15.4% range) and framework design (9.2%) substantially affect performance, that self-evolving skill frameworks can partially close model-capability gaps, and that belief revision difficulty is determined by update design strategy rather than the mere presence of updates. Code is available at https://github.com/aiming-lab/ClawArena.</p> - https://arxiv.org/abs/2604.04202 - Sun, 05 Apr 2026 00:00:00 +0000 - - - Can LLMs Learn to Reason Robustly under Noisy Supervision? - https://arxiv.org/abs/2604.03993 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03993.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shenzhi Yang, Guangcheng Zhu, Bowen Song, Sharon Li, Haobo Wang, Xing Zheng, Yingfan Ma, Zhongqi Chen, Weiqiang Wang, Gang Chen</p><p><b>Upvotes:</b> 41</p><p><b>Summary:</b> Reinforcement Learning with Verifiable Rewards (RLVR) effectively trains reasoning models that rely on abundant perfect labels, but its vulnerability to unavoidable noisy labels due to expert scarcity remains critically underexplored. In this work, we take the first step toward a systematic analysis of noisy label mechanisms in RLVR. In contrast to supervised classification, most RLVR algorithms incorporate a rollout-based condition: a label's influence on training is contingent on whether the current policy can generate rollouts that realize it, a property that naturally extends to noisy labels. Based on this observation, we distinguish two types of noise: inactive noisy labels, which reduce data efficiency, and active noisy labels, which are reinforced and risk skewing the model toward incorrect distributions. From experiments on training with noisy samples, we identify an Early Correctness Coherence phenomenon: although noisy samples begin to lag behind in later stages, accuracy on both clean and noisy samples increases similarly in early training. Motivated by this dynamic, we propose Online Label Refinement (OLR), which progressively corrects potentially noisy labels with majority-voted answers when two conditions hold: a positive slope in the majority answer's rollout pass rate and stable historical consistency across updates, enabling gradual self-correction as the policy improves. We evaluate OLR on six in-distribution mathematical reasoning benchmarks (AIME24/25, AMC, MATH-500, Minerva, and Olympiad) and three out-of-distribution tasks (ARC-c, GPQA-diamond, and MMLU-pro). Across noise ratios from 0.1 to 0.9, OLR consistently improves robustness under both inactive and active noisy-label settings, achieving average gains of 3.6% to 3.9% on in-distribution benchmarks and 3.3% to 4.6% on out-of-distribution evaluations.</p> - https://arxiv.org/abs/2604.03993 - Sun, 05 Apr 2026 00:00:00 +0000 - - - AURA: Always-On Understanding and Real-Time Assistance via Video Streams - https://arxiv.org/abs/2604.04184 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04184.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xudong Lu, Yang Bo, Jinpeng Chen, Shuhan Li, Xintong Guo, Huankang Guan, Fang Liu, Dunyuan Xu, Peiwen Sun, Heyang Sun, Rui Liu, Hongsheng Li</p><p><b>Upvotes:</b> 46</p><p><b>Summary:</b> Video Large Language Models (VideoLLMs) have achieved strong performance on many video understanding tasks, but most existing systems remain offline and are not well-suited for live video streams that require continuous observation and timely response. Recent streaming VideoLLMs have made progress, yet current approaches often rely on decoupled trigger-response pipelines or are limited to captioning-style narration, reducing their effectiveness for open-ended question answering and long-horizon interaction. We propose AURA (Always-On Understanding and Real-Time Assistance), an end-to-end streaming visual interaction framework that enables a unified VideoLLM to continuously process video streams and support both real-time question answering and proactive responses. AURA integrates context management, data construction, training objectives, and deployment optimization for stable long-horizon streaming interaction. It achieves state-of-the-art performance on streaming benchmarks and supports a real-time demo system with ASR and TTS running at 2 FPS on two 80G accelerators. We release the AURA model together with a real-time inference framework to facilitate future research.</p> - https://arxiv.org/abs/2604.04184 - Sun, 05 Apr 2026 00:00:00 +0000 - - - ACES: Who Tests the Tests? Leave-One-Out AUC Consistency for Code Generation - https://arxiv.org/abs/2604.03922 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03922.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hui Sun, Yun-Ji Zhang, Zheng Xie, Ren-Biao Liu, Yali Du, Xin-Ye Li, Ming Li</p><p><b>Upvotes:</b> 52</p><p><b>Summary:</b> Selecting LLM-generated code candidates using LLM-generated tests is challenging because the tests themselves may be incorrect. Existing methods either treat all tests equally or rely on ad-hoc heuristics to filter unreliable tests. Yet determining test correctness requires knowing which codes are correct, creating a circular dependency. Our key insight is that we need not determine test correctness at all: test votes should rank, not merely count. What matters is not how many codes pass a test, but whether the test can distinguish correct from incorrect code. We break the circular dependency via leave-one-out evaluation: hold out one test, rank codes by their aggregate scores on all remaining tests, and measure whether the held-out test's pass/fail pattern agrees with this ranking. We formalize this agreement as the leave-one-out AUC~(LOO-AUC) and prove that the expected LOO-AUC is proportional to each test's ability to separate correct code from incorrect code. Building on this, we propose ACES~(AUC ConsistEncy Scoring) with two complementary variants: ACES-C provides closed-form weights that provably approximate the oracle in expectation under a mild assumption on average test quality; ACES-O drops this assumption and iteratively optimizes a differentiable LOO-AUC objective. Both operate solely on the binary pass matrix with negligible overhead, and achieve state-of-the-art Pass@k on multiple code generation benchmarks.</p> - https://arxiv.org/abs/2604.03922 - Sun, 05 Apr 2026 00:00:00 +0000 - - - Beyond Hard Negatives: The Importance of Score Distribution in Knowledge Distillation for Dense Retrieval - https://arxiv.org/abs/2604.04734 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04734.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Youngjoon Jang, Seongtae Hong, Hyeonseok Moon, Heuiseok Lim</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Transferring knowledge from a cross-encoder teacher via Knowledge Distillation (KD) has become a standard paradigm for training retrieval models. While existing studies have largely focused on mining hard negatives to improve discrimination, the systematic composition of training data and the resulting teacher score distribution have received relatively less attention. In this work, we highlight that focusing solely on hard negatives prevents the student from learning the comprehensive preference structure of the teacher, potentially hampering generalization. To effectively emulate the teacher score distribution, we propose a Stratified Sampling strategy that uniformly covers the entire score spectrum. Experiments on in-domain and out-of-domain benchmarks confirm that Stratified Sampling, which preserves the variance and entropy of teacher scores, serves as a robust baseline, significantly outperforming top-K and random sampling in diverse settings. These findings suggest that the essence of distillation lies in preserving the diverse range of relative scores perceived by the teacher.</p> - https://arxiv.org/abs/2604.04734 + SkVM: Compiling Skills for Efficient Execution Everywhere + https://arxiv.org/abs/2604.03088 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.03088.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Le Chen, Erhu Feng, Yubin Xia, Haibo Chen</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> LLM agents increasingly adopt skills as a reusable unit of composition. While skills are shared across diverse agent platforms, current systems treat them as raw context, causing the same skill to behave inconsistently for different agents. This fragility undermines skill portability and execution efficiency. To address this challenge, we analyze 118,000 skills and draw inspiration from traditional compiler design. We treat skills as code and LLMs as heterogeneous processors. To make portability actionable, we decompose a skill's requirements into a set of primitive capabilities, and measure how well each model-harness pair supports them. Based on these capability profiles, we propose SkVM, a compilation and runtime system designed for portable and efficient skill execution. At compile time, SkVM performs capability-based compilation, environment binding, and concurrency extraction. At runtime, SkVM applies JIT code solidification and adaptive recompilation for performance optimization. We evaluate SkVM across eight LLMs of varying scales and three agent harnesses, covering SkillsBench and representative skill tasks. Results demonstrate that SkVM significantly improves task completion rates across different models and environments while reducing token consumption by up to 40%. In terms of performance, SkVM achieves up to 3.2x speedup with enhanced parallelism, and 19-50x latency reduction through code solidification.</p> + https://arxiv.org/abs/2604.03088 Mon, 06 Apr 2026 00:00:00 +0000 - HDP: A Lightweight Cryptographic Protocol for Human Delegation Provenance in Agentic AI Systems - https://arxiv.org/abs/2604.04522 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04522.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Asiri Dalugoda</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Agentic AI systems increasingly execute consequential actions on behalf of human principals, delegating tasks through multi-step chains of autonomous agents. No existing standard addresses a fundamental accountability gap: verifying that terminal actions in a delegation chain were genuinely authorized by a human principal, through what chain of delegation, and under what scope. This paper presents the Human Delegation Provenance (HDP) protocol, a lightweight token-based scheme that cryptographically captures and verifies human authorization context in multi-agent systems. An HDP token binds a human authorization event to a session, records each agent's delegation action as a signed hop in an append-only chain, and enables any participant to verify the full provenance record using only the issuer's Ed25519 public key and the current session identifier. Verification is fully offline, requiring no registry lookups or third-party trust anchors. We situate HDP within the existing landscape of delegation protocols, identify its distinct design point relative to OAuth 2.0 Token Exchange (RFC 8693), JSON Web Tokens (RFC 7519), UCAN, and the Intent Provenance Protocol (draft-haberkamp-ipp-00), and demonstrate that existing standards fail to address the multi-hop, append-only, human-provenance requirements of agentic systems. HDP has been published as an IETF Internet-Draft (draft-helixar-hdp-agentic-delegation-00) and a reference TypeScript SDK is publicly available.</p> - https://arxiv.org/abs/2604.04522 + Structured Causal Video Reasoning via Multi-Objective Alignment + https://arxiv.org/abs/2604.04415 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04415.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zinuo Li, Yongxin Guo, Jun Liu, Jiawei Zhan, Xi Jiang, Chengjie Wang, Mohammed Bennamoun, Farid Boussaid, Feng Zheng, Qiuhong Ke</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Human understanding of video dynamics is typically grounded in a structured mental representation of entities, actions, and temporal relations, rather than relying solely on immediate deductive reasoning. In contrast, existing Video-LLMs largely depend on unstructured video reasoning, where critical visual evidence is embedded in verbose textual descriptions and temporal causality is often weakly modeled. This leads to inefficient processes and fragile causal inference. To bridge this cognitive gap, we propose constructing a compact representation of salient events and their causal relationships, which we name Structured Event Facts, prior to the reasoning stage. This structured prior serves as an explicit constraint to promote concise and causally grounded reasoning, while also making intermediate evidence easier to verify. To effectively train models on such structured facts, we introduce CausalFact-60K and a four-stage training pipeline comprising facts alignment, format warm-start, thinking warm-start, and reinforcement learning-based post-training. During RL stage, we find that this framework introduces competing objectives, as structural completeness and causal fidelity must be balanced against reasoning length, making it difficult to optimize. We address this challenge by formulating the optimization as a Multi-Objective Reinforcement Learning (MORL) problem and explicitly optimizing toward the Pareto-Frontier to balance these trade-offs. As a result, we introduce Factum-4B, which yields more reliable reasoning and delivers stronger performance on challenging video understanding tasks requiring fine-grained temporal inference.</p> + https://arxiv.org/abs/2604.04415 Mon, 06 Apr 2026 00:00:00 +0000 - CLEAR: Unlocking Generative Potential for Degraded Image Understanding in Unified Multimodal Models - https://arxiv.org/abs/2604.04780 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04780.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiangzhao Hao, Zefeng Zhang, Zhenyu Zhang, Linhao Yu, Yao Chen, Yiqian Zhang, Haiyun Guo, Shuohuan Wang, Yu Sun</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Image degradation from blur, noise, compression, and poor illumination severely undermines multimodal understanding in real-world settings. Unified multimodal models that combine understanding and generation within a single architecture are a natural fit for this challenge, as their generative pathway can model the fine-grained visual structure that degradation destroys. Yet these models fail to leverage their own generative capacity on degraded inputs. We trace this disconnect to two compounding factors: existing training regimes never ask the model to invoke generation during reasoning, and the standard decode-reencode pathway does not support effective joint optimization. We present CLEAR, a framework that connects the two capabilities through three progressive steps: (1) supervised fine-tuning on a degradation-aware dataset to establish the generate-then-answer reasoning pattern; (2) a Latent Representation Bridge that replaces the decode-reencode detour with a direct, optimizable connection between generation and reasoning; (3) Interleaved GRPO, a reinforcement learning method that jointly optimizes text reasoning and visual generation under answer-correctness rewards. We construct MMD-Bench, covering three degradation severity levels across six standard multimodal benchmarks. Experiments show that CLEAR substantially improves robustness on degraded inputs while preserving clean-image performance. Our analysis further reveals that removing pixel-level reconstruction supervision leads to intermediate visual states with higher perceptual quality, suggesting that task-driven optimization and visual quality are naturally aligned.</p> - https://arxiv.org/abs/2604.04780 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Paper Espresso: From Paper Overload to Research Insight - https://arxiv.org/abs/2604.04562 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04562.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingzhe Du, Luu Anh Tuan, Dong Huang, See-kiong Ng</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> The accelerating pace of scientific publishing makes it increasingly difficult for researchers to stay current. We present Paper Espresso, an open-source platform that automatically discovers, summarizes, and analyzes trending arXiv papers. The system uses large language models (LLMs) to generate structured summaries with topical labels and keywords, and provides multi-granularity trend analysis at daily, weekly, and monthly scales through LLM-driven topic consolidation. Over 35 months of continuous deployment, Paper Espresso has processed over 13,300 papers and publicly released all structured metadata, revealing rich dynamics in the AI research landscape: a mid-2025 surge in reinforcement learning for LLM reasoning, non-saturating topic emergence (6,673 unique topics), and a positive correlation between topic novelty and community engagement (2.0x median upvotes for the most novel papers). A live demo is available at https://huggingface.co/spaces/Elfsong/Paper_Espresso.</p> - https://arxiv.org/abs/2604.04562 - Mon, 06 Apr 2026 00:00:00 +0000 - - - AvatarPointillist: AutoRegressive 4D Gaussian Avatarization - https://arxiv.org/abs/2604.04787 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04787.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hongyu Liu, Xuan Wang, Yating Wang, Zijian Wu, Ziyu Wan, Yue Ma, Runtao Liu, Boyao Zhou, Yujun Shen, Qifeng Chen</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> We introduce AvatarPointillist, a novel framework for generating dynamic 4D Gaussian avatars from a single portrait image. At the core of our method is a decoder-only Transformer that autoregressively generates a point cloud for 3D Gaussian Splatting. This sequential approach allows for precise, adaptive construction, dynamically adjusting point density and the total number of points based on the subject's complexity. During point generation, the AR model also jointly predicts per-point binding information, enabling realistic animation. After generation, a dedicated Gaussian decoder converts the points into complete, renderable Gaussian attributes. We demonstrate that conditioning the decoder on the latent features from the AR generator enables effective interaction between stages and markedly improves fidelity. Extensive experiments validate that AvatarPointillist produces high-quality, photorealistic, and controllable avatars. We believe this autoregressive formulation represents a new paradigm for avatar generation, and we will release our code inspire future research.</p> - https://arxiv.org/abs/2604.04787 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Less Detail, Better Answers: Degradation-Driven Prompting for VQA - https://arxiv.org/abs/2604.04838 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04838.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoxuan Han, Weijie Wang, Zeyu Zhang, Yefei He, Bohan Zhuang</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Recent advancements in Vision-Language Models (VLMs) have significantly pushed the boundaries of Visual Question Answering (VQA).However,high-resolution details can sometimes become noise that leads to hallucinations or reasoning errors. In this paper,we propose Degradation-Driven Prompting (DDP), a novel framework that improves VQA performance by strategically reducing image fidelity to force models to focus on essential structural information. We evaluate DDP across two distinct tasks. Physical attributes targets images prone to human misjudgment, where DDP employs a combination of 80p downsampling, structural visual aids (white background masks and orthometric lines), and In-Context Learning (ICL) to calibrate the model's focus. Perceptual phenomena addresses various machine-susceptible visual anomalies and illusions, including Visual Anomaly (VA), Color (CI), Motion(MI),Gestalt (GI), Geometric (GSI), and Visual Illusions (VI).For this task, DDP integrates a task-classification stage with specialized tools such as blur masks and contrast enhancement alongside downsampling. Our experimental results demonstrate that less is more: by intentionally degrading visual inputs and providing targeted structural prompts, DDP enables VLMs to bypass distracting textures and achieve superior reasoning accuracy on challenging visual benchmarks.</p> - https://arxiv.org/abs/2604.04838 - Mon, 06 Apr 2026 00:00:00 +0000 - - - MedGemma 1.5 Technical Report - https://arxiv.org/abs/2604.05081 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05081.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Andrew Sellergren, Chufan Gao, Fereshteh Mahvar, Timo Kohlberger, Fayaz Jamil, Madeleine Traverse, Alberto Tono, Bashir Sadjad, Lin Yang, Charles Lau, Liron Yatziv, Tiffany Chen, Bram Sterling, Kenneth Philbrick, Richa Tiwari, Yun Liu, Madhuram Jajoo, Chandrashekar Sankarapu, Swapnil Vispute, Harshad Purandare, Abhishek Bijay Mishra, Sam Schmidgall, Tao Tu, Anil Palepu, Chunjong Park, Tim Strother, Rahul Thapa, Yong Cheng, Preeti Singh, Kat Black, Yossi Matias, Katherine Chou, Avinatan Hassidim, Kavi Goel, Joelle Barral, Tris Warkentin, Shravya Shetty, Dale Webster, Sunny Virmani, David F. Steiner, Can Kirmizibayrak, Daniel Golden</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> We introduce MedGemma 1.5 4B, the latest model in the MedGemma collection. MedGemma 1.5 expands on MedGemma 1 by integrating additional capabilities: high-dimensional medical imaging (CT/MRI volumes and histopathology whole slide images), anatomical localization via bounding boxes, multi-timepoint chest X-ray analysis, and improved medical document understanding (lab reports, electronic health records). We detail the innovations required to enable these modalities within a single architecture, including new training data, long-context 3D volume slicing, and whole-slide pathology sampling. Compared to MedGemma 1 4B, MedGemma 1.5 4B demonstrates significant gains in these new areas, improving 3D MRI condition classification accuracy by 11% and 3D CT condition classification by 3% (absolute improvements). In whole slide pathology imaging, MedGemma 1.5 4B achieves a 47% macro F1 gain. Additionally, it improves anatomical localization with a 35% increase in Intersection over Union on chest X-rays and achieves a 4% macro accuracy for longitudinal (multi-timepoint) chest x-ray analysis. Beyond its improved multimodal performance over MedGemma 1, MedGemma 1.5 improves on text-based clinical knowledge and reasoning, improving by 5% on MedQA accuracy and 22% on EHRQA accuracy. It also achieves an average of 18% macro F1 on 4 different lab report information extraction datasets (EHR Datasets 2, 3, 4, and Mendeley Clinical Laboratory Test Reports). Taken together, MedGemma 1.5 serves as a robust, open resource for the community, designed as an improved foundation on which developers can create the next generation of medical AI systems. Resources and tutorials for building upon MedGemma 1.5 can be found at https://goo.gle/MedGemma.</p> - https://arxiv.org/abs/2604.05081 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Synthetic Sandbox for Training Machine Learning Engineering Agents - https://arxiv.org/abs/2604.04872 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04872.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuhang Zhou, Lizhu Zhang, Yifan Wu, Jiayi Liu, Xiangjun Fan, Zhuokai Zhao, Hong Yan</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> As large language model agents advance beyond software engineering (SWE) tasks toward machine learning engineering (MLE), verifying agent behavior becomes orders of magnitude more expensive: while SWE tasks can be verified via fast-executing unit tests, MLE verification requires running full ML pipelines -- data preprocessing, model training, and metric evaluation -- on large datasets at each rollout step, rendering trajectory-wise on-policy reinforcement learning (RL) prohibitively slow. Existing approaches retreat to supervised fine-tuning (SFT) or offline proxy rewards, sacrificing the exploration and generalization benefits of on-policy RL. We observe that sandbox data size is the primary source of this bottleneck. Based on this insight, we introduce SandMLE, a multi-agent framework that generates diverse, verifiable synthetic MLE environments from a small number of seed tasks, preserving the structural and technical complexity of real-world problems while constraining datasets to micro-scale (each task is paired with only 50-200 training samples). Through extensive experiments, we show that SandMLE reduces execution time by over 13 times, enabling large-scale, on-policy trajectory-wise RL for the first time in the MLE domain. On MLE-bench-lite, SandMLE yields significant gains over SFT baselines across Qwen3-8B, 14B, and 30B-A3B, with relative medal rate improvements ranging from 20.3% to 66.9%. Furthermore, the trained policy generalizes across unseen agentic scaffolds, achieving up to 32.4% better HumanRank score on MLE-Dojo.</p> - https://arxiv.org/abs/2604.04872 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Demystifying When Pruning Works via Representation Hierarchies - https://arxiv.org/abs/2603.24652 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2603.24652.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shwai He, Guoheng Sun, Haichao Zhang, Yun Fu, Ang Li</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Network pruning, which removes less important parameters or architectures, is often expected to improve efficiency while preserving performance. However, this expectation does not consistently hold across language tasks: pruned models can perform well on non-generative tasks but frequently fail in generative settings. To understand this discrepancy, we analyze network pruning from a representation-hierarchy perspective, decomposing the internal computation of language models into three sequential spaces: embedding (hidden representations), logit (pre-softmax outputs), and probability (post-softmax distributions). We find that representations in the embedding and logit spaces are largely robust to pruning-induced perturbations. However, the nonlinear transformation from logits to probabilities amplifies these deviations, which accumulate across time steps and lead to substantial degradation during generation. In contrast, the stability of the categorical-token probability subspace, together with the robustness of the embedding space, supports the effectiveness of pruning for non-generative tasks such as retrieval and multiple-choice selection. Our analysis disentangles the effects of pruning across tasks and provides practical guidance for its application. Code is available at https://github.com/CASE-Lab-UMD/Pruning-on-Representations</p> - https://arxiv.org/abs/2603.24652 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw - https://arxiv.org/abs/2604.04759 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04759.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zijun Wang, Haoqin Tu, Letian Zhang, Hardy Chen, Juncheng Wu, Xiangyan Liu, Zhenlong Yuan, Tianyu Pang, Michael Qizhe Shieh, Fengze Liu, Zeyu Zheng, Huaxiu Yao, Yuyin Zhou, Cihang Xie</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> OpenClaw, the most widely deployed personal AI agent in early 2026, operates with full local system access and integrates with sensitive services such as Gmail, Stripe, and the filesystem. While these broad privileges enable high levels of automation and powerful personalization, they also expose a substantial attack surface that existing sandboxed evaluations fail to capture. To address this gap, we present the first real-world safety evaluation of OpenClaw and introduce the CIK taxonomy, which unifies an agent's persistent state into three dimensions, i.e., Capability, Identity, and Knowledge, for safety analysis. Our evaluations cover 12 attack scenarios on a live OpenClaw instance across four backbone models (Claude Sonnet 4.5, Opus 4.6, Gemini 3.1 Pro, and GPT-5.4). The results show that poisoning any single CIK dimension increases the average attack success rate from 24.6% to 64-74%, with even the most robust model exhibiting more than a threefold increase over its baseline vulnerability. We further assess three CIK-aligned defense strategies alongside a file-protection mechanism; however, the strongest defense still yields a 63.8% success rate under Capability-targeted attacks, while file protection blocks 97% of malicious injections but also prevents legitimate updates. Taken together, these findings show that the vulnerabilities are inherent to the agent architecture, necessitating more systematic safeguards to secure personal AI agents. Our project page is https://ucsc-vlaa.github.io/CIK-Bench.</p> - https://arxiv.org/abs/2604.04759 - Mon, 06 Apr 2026 00:00:00 +0000 - - - ClawsBench: Evaluating Capability and Safety of LLM Productivity Agents in Simulated Workspaces - https://arxiv.org/abs/2604.05172 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05172.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiangyi Li, Kyoung Whan Choe, Yimin Liu, Xiaokun Chen, Chujun Tao, Bingran You, Wenbo Chen, Zonglin Di, Jiankai Sun, Shenghan Zheng, Jiajun Bao, Yuanli Wang, Weixiang Yan, Yiyuan Li, Han-chung Lee</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> Large language model (LLM) agents are increasingly deployed to automate productivity tasks (e.g., email, scheduling, document management), but evaluating them on live services is risky due to potentially irreversible changes. Existing benchmarks rely on simplified environments and fail to capture realistic, stateful, multi-service workflows. We introduce ClawsBench, a benchmark for evaluating and improving LLM agents in realistic productivity settings. It includes five high-fidelity mock services (Gmail, Slack, Google Calendar, Google Docs, Google Drive) with full state management and deterministic snapshot/restore, along with 44 structured tasks covering single-service, cross-service, and safety-critical scenarios. We decompose agent scaffolding into two independent levers (domain skills that inject API knowledge via progressive disclosure, and a meta prompt that coordinates behavior across services) and vary both to measure their separate and combined effects. Experiments across 6 models, 4 agent harnesses, and 33 conditions show that with full scaffolding, agents achieve task success rates of 39-64% but exhibit unsafe action rates of 7-33%. On OpenClaw, the top five models fall within a 10 percentage-point band on task success (53-63%), with unsafe action rates from 7% to 23% and no consistent ordering between the two metrics. We identify eight recurring patterns of unsafe behavior, including multi-step sandbox escalation and silent contract modification.</p> - https://arxiv.org/abs/2604.05172 - Mon, 06 Apr 2026 00:00:00 +0000 - - - General Multimodal Protein Design Enables DNA-Encoding of Chemistry - https://arxiv.org/abs/2604.05181 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05181.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jarrid Rector-Brooks, Théophile Lambert, Marta Skreta, Daniel Roth, Yueming Long, Zi-Qi Li, Xi Zhang, Miruna Cretu, Francesca-Zhoufan Li, Tanvi Ganapathy, Emily Jin, Avishek Joey Bose, Jason Yang, Kirill Neklyudov, Yoshua Bengio, Alexander Tong, Frances H. Arnold, Cheng-Hao Liu</p><p><b>Upvotes:</b> 29</p><p><b>Summary:</b> Evolution is an extraordinary engine for enzymatic diversity, yet the chemistry it has explored remains a narrow slice of what DNA can encode. Deep generative models can design new proteins that bind ligands, but none have created enzymes without pre-specifying catalytic residues. We introduce DISCO (DIffusion for Sequence-structure CO-design), a multimodal model that co-designs protein sequence and 3D structure around arbitrary biomolecules, as well as inference-time scaling methods that optimize objectives across both modalities. Conditioned solely on reactive intermediates, DISCO designs diverse heme enzymes with novel active-site geometries. These enzymes catalyze new-to-nature carbene-transfer reactions, including alkene cyclopropanation, spirocyclopropanation, B-H, and C(sp^3)-H insertions, with high activities exceeding those of engineered enzymes. Random mutagenesis of a selected design further confirmed that enzyme activity can be improved through directed evolution. By providing a scalable route to evolvable enzymes, DISCO broadens the potential scope of genetically encodable transformations. Code is available at https://github.com/DISCO-design/DISCO.</p> - https://arxiv.org/abs/2604.05181 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Vero: An Open RL Recipe for General Visual Reasoning - https://arxiv.org/abs/2604.04917 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04917.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Gabriel Sarch, Linrong Cai, Qunzhong Wang, Haoyang Wu, Danqi Chen, Zhuang Liu</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> What does it take to build a visual reasoner that works across charts, science, spatial understanding, and open-ended tasks? The strongest vision-language models (VLMs) show such broad visual reasoning is within reach, but the recipe behind them remains unclear, locked behind proprietary reinforcement learning (RL) pipelines with non-public data. We introduce Vero, a family of fully open VLMs that matches or exceeds existing open-weight models across diverse visual reasoning tasks. We scale RL data and rewards across six broad task categories, constructing Vero-600K, a 600K-sample dataset from 59 datasets, and designing task-routed rewards that handle heterogeneous answer formats. Vero achieves state-of-the-art performance, improving over four base models by 3.7-5.5 points on average across VeroEval, our suite of 30 challenging benchmarks. Starting from Qwen3-VL-8B-Instruct, Vero outperforms Qwen3-VL-8B-Thinking on 23 of 30 benchmarks without additional proprietary thinking data. When trained from the same base model, Vero-600K exceeds existing RL datasets across task categories. Systematic ablations reveal that different task categories elicit qualitatively distinct reasoning patterns that transfer poorly in isolation, suggesting that broad data coverage is the primary driver of strong RL scaling. All data, code, and models are released.</p> - https://arxiv.org/abs/2604.04917 - Mon, 06 Apr 2026 00:00:00 +0000 - - - SkillX: Automatically Constructing Skill Knowledge Bases for Agents - https://arxiv.org/abs/2604.04804 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04804.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenxi Wang, Zhuoyun Yu, Xin Xie, Wuguannan Yao, Runnan Fang, Shuofei Qiao, Kexin Cao, Guozhou Zheng, Xiang Qi, Peng Zhang, Shumin Deng</p><p><b>Upvotes:</b> 31</p><p><b>Summary:</b> Learning from experience is critical for building capable large language model (LLM) agents, yet prevailing self-evolving paradigms remain inefficient: agents learn in isolation, repeatedly rediscover similar behaviors from limited experience, resulting in redundant exploration and poor generalization. To address this problem, we propose SkillX, a fully automated framework for constructing a plug-and-play skill knowledge base that can be reused across agents and environments. SkillX operates through a fully automated pipeline built on three synergistic innovations: (i) Multi-Level Skills Design, which distills raw trajectories into three-tiered hierarchy of strategic plans, functional skills, and atomic skills; (ii) Iterative Skills Refinement, which automatically revises skills based on execution feedback to continuously improve library quality; and (iii) Exploratory Skills Expansion, which proactively generates and validates novel skills to expand coverage beyond seed training data. Using a strong backbone agent (GLM-4.6), we automatically build a reusable skill library and evaluate its transferability on challenging long-horizon, user-interactive benchmarks, including AppWorld, BFCL-v3, and τ^2-Bench. Experiments show that SkillKB consistently improves task success and execution efficiency when plugged into weaker base agents, highlighting the importance of structured, hierarchical experience representations for generalizable agent learning. Our code will be publicly available soon at https://github.com/zjunlp/SkillX.</p> - https://arxiv.org/abs/2604.04804 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Watch Before You Answer: Learning from Visually Grounded Post-Training - https://arxiv.org/abs/2604.05117 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05117.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuxuan Zhang, EunJeong Hwang, Huaisong Zhang, Penghui Du, Yiming Jia, Dongfu Jiang, Xuan He, Shenhui Zhang, Ping Nie, Peter West, Kelsey R. Allen</p><p><b>Upvotes:</b> 34</p><p><b>Summary:</b> It is critical for vision-language models (VLMs) to comprehensively understand visual, temporal, and textual cues. However, despite rapid progress in multimodal modeling, video understanding performance still lags behind text-based reasoning. In this work, we find that progress is even worse than previously assumed: commonly reported long video understanding benchmarks contain 40-60% of questions that can be answered using text cues alone. Furthermore, we find that these issues are also pervasive in widely used post-training datasets, potentially undercutting the ability of post-training to improve VLM video understanding performance. Guided by this observation, we introduce VidGround as a simple yet effective solution: using only the actual visually grounded questions without any linguistic biases for post-training. When used in tandem with RL-based post-training algorithms, this simple technique improves performance by up to 6.2 points relative to using the full dataset, while using only 69.1% of the original post-training data. Moreover, we show that data curation with a simple post-training algorithm outperforms several more complex post-training techniques, highlighting that data quality is a major bottleneck for improving video understanding in VLMs. These results underscore the importance of curating post-training data and evaluation benchmarks that truly require visual grounding to advance the development of more capable VLMs. Project page: http://vidground.etuagi.com.</p> - https://arxiv.org/abs/2604.05117 - Mon, 06 Apr 2026 00:00:00 +0000 - - - SpatialEdit: Benchmarking Fine-Grained Image Spatial Editing - https://arxiv.org/abs/2604.04911 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04911.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yicheng Xiao, Wenhu Zhang, Lin Song, Yukang Chen, Wenbo Li, Nan Jiang, Tianhe Ren, Haokun Lin, Wei Huang, Haoyang Huang, Xiu Li, Nan Duan, Xiaojuan Qi</p><p><b>Upvotes:</b> 35</p><p><b>Summary:</b> Image spatial editing performs geometry-driven transformations, allowing precise control over object layout and camera viewpoints. Current models are insufficient for fine-grained spatial manipulations, motivating a dedicated assessment suite. Our contributions are listed: (i) We introduce SpatialEdit-Bench, a complete benchmark that evaluates spatial editing by jointly measuring perceptual plausibility and geometric fidelity via viewpoint reconstruction and framing analysis. (ii) To address the data bottleneck for scalable training, we construct SpatialEdit-500k, a synthetic dataset generated with a controllable Blender pipeline that renders objects across diverse backgrounds and systematic camera trajectories, providing precise ground-truth transformations for both object- and camera-centric operations. (iii) Building on this data, we develop SpatialEdit-16B, a baseline model for fine-grained spatial editing. Our method achieves competitive performance on general editing while substantially outperforming prior methods on spatial manipulation tasks. All resources will be made public at https://github.com/EasonXiao-888/SpatialEdit.</p> - https://arxiv.org/abs/2604.04911 - Mon, 06 Apr 2026 00:00:00 +0000 - - - How Well Do Agentic Skills Work in the Wild: Benchmarking LLM Skill Usage in Realistic Settings - https://arxiv.org/abs/2604.04323 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04323.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yujian Liu, Jiabao Ji, Li An, Tommi Jaakkola, Yang Zhang, Shiyu Chang</p><p><b>Upvotes:</b> 38</p><p><b>Summary:</b> Agent skills, which are reusable, domain-specific knowledge artifacts, have become a popular mechanism for extending LLM-based agents, yet formally benchmarking skill usage performance remains scarce. Existing skill benchmarking efforts focus on overly idealized conditions, where LLMs are directly provided with hand-crafted, narrowly-tailored task-specific skills for each task, whereas in many realistic settings, the LLM agent may have to search for and select relevant skills on its own, and even the closest matching skills may not be well-tailored for the task. In this paper, we conduct the first comprehensive study of skill utility under progressively challenging realistic settings, where agents must retrieve skills from a large collection of 34k real-world skills and may not have access to any hand-curated skills. Our findings reveal that the benefits of skills are fragile: performance gains degrade consistently as settings become more realistic, with pass rates approaching no-skill baselines in the most challenging scenarios. To narrow this gap, we study skill refinement strategies, including query-specific and query-agnostic approaches, and we show that query-specific refinement substantially recovers lost performance when the initial skills are of reasonable relevance and quality. We further demonstrate the generality of retrieval and refinement on Terminal-Bench 2.0, where they improve the pass rate of Claude Opus 4.6 from 57.7% to 65.5%. Our results, consistent across multiple models, highlight both the promise and the current limitations of skills for LLM-based agents. Our code is available at https://github.com/UCSB-NLP-Chang/Skill-Usage.</p> - https://arxiv.org/abs/2604.04323 - Mon, 06 Apr 2026 00:00:00 +0000 - - - FileGram: Grounding Agent Personalization in File-System Behavioral Traces - https://arxiv.org/abs/2604.04901 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04901.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuai Liu, Shulin Tian, Kairui Hu, Yuhao Dong, Zhe Yang, Bo Li, Jingkang Yang, Chen Change Loy, Ziwei Liu</p><p><b>Upvotes:</b> 40</p><p><b>Summary:</b> Coworking AI agents operating within local file systems are rapidly emerging as a paradigm in human-AI interaction; however, effective personalization remains limited by severe data constraints, as strict privacy barriers and the difficulty of jointly collecting multimodal real-world traces prevent scalable training and evaluation, and existing methods remain interaction-centric while overlooking dense behavioral traces in file-system operations; to address this gap, we propose FileGram, a comprehensive framework that grounds agent memory and personalization in file-system behavioral traces, comprising three core components: (1) FileGramEngine, a scalable persona-driven data engine that simulates realistic workflows and generates fine-grained multimodal action sequences at scale; (2) FileGramBench, a diagnostic benchmark grounded in file-system behavioral traces for evaluating memory systems on profile reconstruction, trace disentanglement, persona drift detection, and multimodal grounding; and (3) FileGramOS, a bottom-up memory architecture that builds user profiles directly from atomic actions and content deltas rather than dialogue summaries, encoding these traces into procedural, semantic, and episodic channels with query-time abstraction; extensive experiments show that FileGramBench remains challenging for state-of-the-art memory systems and that FileGramEngine and FileGramOS are effective, and by open-sourcing the framework, we hope to support future research on personalized memory-centric file-system agents.</p> - https://arxiv.org/abs/2604.04901 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Vanast: Virtual Try-On with Human Image Animation via Synthetic Triplet Supervision - https://arxiv.org/abs/2604.04934 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04934.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hyunsoo Cha, Wonjung Woo, Byungjun Kim, Hanbyul Joo</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> We present Vanast, a unified framework that generates garment-transferred human animation videos directly from a single human image, garment images, and a pose guidance video. Conventional two-stage pipelines treat image-based virtual try-on and pose-driven animation as separate processes, which often results in identity drift, garment distortion, and front-back inconsistency. Our model addresses these issues by performing the entire process in a single unified step to achieve coherent synthesis. To enable this setting, we construct large-scale triplet supervision. Our data generation pipeline includes generating identity-preserving human images in alternative outfits that differ from garment catalog images, capturing full upper and lower garment triplets to overcome the single-garment-posed video pair limitation, and assembling diverse in-the-wild triplets without requiring garment catalog images. We further introduce a Dual Module architecture for video diffusion transformers to stabilize training, preserve pretrained generative quality, and improve garment accuracy, pose adherence, and identity preservation while supporting zero-shot garment interpolation. Together, these contributions allow Vanast to produce high-fidelity, identity-consistent animation across a wide range of garment types.</p> - https://arxiv.org/abs/2604.04934 - Mon, 06 Apr 2026 00:00:00 +0000 - - - MegaTrain: Full Precision Training of 100B+ Parameter Large Language Models on a Single GPU - https://arxiv.org/abs/2604.05091 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05091.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhengqing Yuan, Hanchi Sun, Lichao Sun, Yanfang Ye</p><p><b>Upvotes:</b> 43</p><p><b>Summary:</b> We present MegaTrain, a memory-centric system that efficiently trains 100B+ parameter large language models at full precision on a single GPU. Unlike traditional GPU-centric systems, MegaTrain stores parameters and optimizer states in host memory (CPU memory) and treats GPUs as transient compute engines. For each layer, we stream parameters in and compute gradients out, minimizing persistent device state. To battle the CPU-GPU bandwidth bottleneck, we adopt two key optimizations. 1) We introduce a pipelined double-buffered execution engine that overlaps parameter prefetching, computation, and gradient offloading across multiple CUDA streams, enabling continuous GPU execution. 2) We replace persistent autograd graphs with stateless layer templates, binding weights dynamically as they stream in, eliminating persistent graph metadata while providing flexibility in scheduling. On a single H200 GPU with 1.5TB host memory, MegaTrain reliably trains models up to 120B parameters. It also achieves 1.84times the training throughput of DeepSpeed ZeRO-3 with CPU offloading when training 14B models. MegaTrain also enables 7B model training with 512k token context on a single GH200.</p> - https://arxiv.org/abs/2604.05091 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Memory Intelligence Agent - https://arxiv.org/abs/2604.04503 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04503.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jingyang Qiao, Weicheng Meng, Yu Cheng, Zhihang Lin, Zhizhong Zhang, Xin Tan, Jingyu Gong, Kun Shao, Yuan Xie</p><p><b>Upvotes:</b> 52</p><p><b>Summary:</b> Deep research agents (DRAs) integrate LLM reasoning with external tools. Memory systems enable DRAs to leverage historical experiences, which are essential for efficient reasoning and autonomous evolution. Existing methods rely on retrieving similar trajectories from memory to aid reasoning, while suffering from key limitations of ineffective memory evolution and increasing storage and retrieval costs. To address these problems, we propose a novel Memory Intelligence Agent (MIA) framework, consisting of a Manager-Planner-Executor architecture. Memory Manager is a non-parametric memory system that can store compressed historical search trajectories. Planner is a parametric memory agent that can produce search plans for questions. Executor is another agent that can search and analyze information guided by the search plan. To build the MIA framework, we first adopt an alternating reinforcement learning paradigm to enhance cooperation between the Planner and the Executor. Furthermore, we enable the Planner to continuously evolve during test-time learning, with updates performed on-the-fly alongside inference without interrupting the reasoning process. Additionally, we establish a bidirectional conversion loop between parametric and non-parametric memories to achieve efficient memory evolution. Finally, we incorporate a reflection and an unsupervised judgment mechanisms to boost reasoning and self-evolution in the open world. Extensive experiments across eleven benchmarks demonstrate the superiority of MIA.</p> - https://arxiv.org/abs/2604.04503 - Mon, 06 Apr 2026 00:00:00 +0000 - - - TriAttention: Efficient Long Reasoning with Trigonometric KV Compression - https://arxiv.org/abs/2604.04921 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04921.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Weian Mao, Xi Lin, Wei Huang, Yuxin Xie, Tianfu Fu, Bohan Zhuang, Song Han, Yukang Chen</p><p><b>Upvotes:</b> 103</p><p><b>Summary:</b> Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks. Leading KV cache compression methods estimate KV importance using attention scores from recent post-RoPE queries. However, queries rotate with position during RoPE, making representative queries very few, leading to poor top-key selection and unstable reasoning. To avoid this issue, we turn to the pre-RoPE space, where we observe that Q and K vectors are highly concentrated around fixed non-zero centers and remain stable across positions -- Q/K concentration. We show that this concentration causes queries to preferentially attend to keys at specific distances (e.g., nearest keys), with the centers determining which distances are preferred via a trigonometric series. Based on this, we propose TriAttention to estimate key importance by leveraging these centers. Via the trigonometric series, we use the distance preference characterized by these centers to score keys according to their positions, and also leverage Q/K norms as an additional signal for importance estimation. On AIME25 with 32K-token generation, TriAttention matches Full Attention reasoning accuracy while achieving 2.5x higher throughput or 10.7x KV memory reduction, whereas leading baselines achieve only about half the accuracy at the same efficiency. TriAttention enables OpenClaw deployment on a single consumer GPU, where long context would otherwise cause out-of-memory with Full Attention.</p> - https://arxiv.org/abs/2604.04921 - Mon, 06 Apr 2026 00:00:00 +0000 - - - MinerU2.5-Pro: Pushing the Limits of Data-Centric Document Parsing at Scale - https://arxiv.org/abs/2604.04771 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04771.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bin Wang, Tianyao He, Linke Ouyang, Fan Wu, Zhiyuan Zhao, Tao Chu, Yuan Qu, Zhenjiang Jin, Weijun Zeng, Ziyang Miao, Bangrui Xu, Junbo Niu, Mengzhang Cai, Jiantao Qiu, Qintong Zhang, Dongsheng Ma, Yuefeng Sun, Hejun Dong, Wenzheng Zhang, Jutao Xiao, Jiayong Shi, Pengyu Liao, Xiaomeng Zhao, Huaping Zhong, Liqun Wei, Jing Yu, Jie Yang, Wei Li, Shasha Wang, Qianqian Wu, Xuanhe Zhou, Weijia Li, Zhenxiang Li, Zhongying Tu, Jiang Wu, Lijun Wu, Chao Xu, Kai Chen, Wentao Zhang, Yu Qiao, Bowen Zhou, Dahua Lin, Conghui He</p><p><b>Upvotes:</b> 115</p><p><b>Summary:</b> Current document parsing methods compete primarily on model architecture innovation, while systematic engineering of training data remains underexplored. Yet SOTA models of different architectures and parameter scales exhibit highly consistent failure patterns on the same set of hard samples, suggesting that the performance bottleneck stems from shared deficiencies in training data rather than architecture itself. Building on this finding, we present \minerupro, which advances the state of the art solely through data engineering and training strategy optimization while keeping the 1.2B-parameter architecture of \mineru completely fixed. At its core is a Data Engine co-designed around coverage, informativeness, and annotation accuracy: Diversity-and-Difficulty-Aware Sampling expands training data from under 10M to 65.5M samples while correcting distribution shift; Cross-Model Consistency Verification leverages output agreement among heterogeneous models to assess sample difficulty and generate reliable annotations; the Judge-and-Refine pipeline improves annotation quality for hard samples through render-then-verify iterative correction. A three-stage progressive training strategy -- large-scale pre-training, hard sample fine-tuning, and GRPO alignment -- sequentially exploits these data at different quality tiers. On the evaluation front, we fix element-matching biases in OmniDocBench~v1.5 and introduce a Hard subset, establishing the more discriminative OmniDocBench~v1.6 protocol. Without any architectural modification, \minerupro achieves 95.69 on OmniDocBench~v1.6, improving over the same-architecture baseline by 2.71 points and surpassing all existing methods including models with over 200times more parameters.</p> - https://arxiv.org/abs/2604.04771 - Mon, 06 Apr 2026 00:00:00 +0000 - - - OpenWorldLib: A Unified Codebase and Definition of Advanced World Models - https://arxiv.org/abs/2604.04707 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04707.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> DataFlow Team, Bohan Zeng, Daili Hua, Kaixin Zhu, Yifan Dai, Bozhou Li, Yuran Wang, Chengzhuo Tong, Yifan Yang, Mingkun Chang, Jianbin Zhao, Zhou Liu, Hao Liang, Xiaochen Ma, Ruichuan An, Junbo Niu, Zimo Meng, Tianyi Bai, Meiyi Qiang, Huanyao Zhang, Zhiyou Xiao, Tianyu Guo, Qinhan Yu, Runhao Zhao, Zhengpin Li, Xinyi Huang, Yisheng Pan, Yiwen Tang, Yang Shi, Yue Ding, Xinlong Chen, Hongcheng Gao, Minglei Shi, Jialong Wu, Zekun Wang, Yuanxing Zhang, Xintao Wang, Pengfei Wan, Yiren Song, Mike Zheng Shou, Wentao Zhang</p><p><b>Upvotes:</b> 200</p><p><b>Summary:</b> World models have garnered significant attention as a promising research direction in artificial intelligence, yet a clear and unified definition remains lacking. In this paper, we introduce OpenWorldLib, a comprehensive and standardized inference framework for Advanced World Models. Drawing on the evolution of world models, we propose a clear definition: a world model is a model or framework centered on perception, equipped with interaction and long-term memory capabilities, for understanding and predicting the complex world. We further systematically categorize the essential capabilities of world models. Based on this definition, OpenWorldLib integrates models across different tasks within a unified framework, enabling efficient reuse and collaborative inference. Finally, we present additional reflections and analyses on potential future directions for world model research. Code link: https://github.com/OpenDCAI/OpenWorldLib</p> - https://arxiv.org/abs/2604.04707 - Mon, 06 Apr 2026 00:00:00 +0000 - - - Video-MME-v2: Towards the Next Stage in Benchmarks for Comprehensive Video Understanding - https://arxiv.org/abs/2604.05015 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05015.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chaoyou Fu, Haozhi Yuan, Yuhao Dong, Yi-Fan Zhang, Yunhang Shen, Xiaoxing Hu, Xueying Li, Jinsen Su, Chengwu Long, Xiaoyao Xie, Yongkang Xie, Xiawu Zheng, Xue Yang, Haoyu Cao, Yunsheng Wu, Ziwei Liu, Xing Sun, Caifeng Shan, Ran He</p><p><b>Upvotes:</b> 227</p><p><b>Summary:</b> With the rapid advancement of video understanding, existing benchmarks are becoming increasingly saturated, exposing a critical discrepancy between inflated leaderboard scores and real-world model capabilities. To address this widening gap, we introduce Video-MME-v2, a comprehensive benchmark designed to rigorously evaluate the robustness and faithfulness of video understanding. To systematically evaluate model capabilities, we design a progressive tri-level hierarchy that incrementally increases the complexity of video comprehension, ranging from multi-point visual information aggregation, to temporal dynamics modeling, and ultimately to complex multimodal reasoning. Besides, in contrast to conventional per-question accuracy, we propose a group-based non-linear evaluation strategy that enforces both consistency across related queries and coherence in multi-step reasoning. It penalizes fragmented or guess-based correctness and assigns credit only to answers supported by valid reasoning. To guarantee data quality, Video-MME-v2 is constructed through a rigorously controlled human annotation pipeline, involving 12 annotators and 50 independent reviewers. Backed by 3,300 human-hours and up to 5 rounds of quality assurance, Video-MME-v2 aims to serve as one of the most authoritative video benchmarks. Extensive experiments reveal a substantial gap between current best model Gemini-3-Pro and human experts, and uncover a clear hierarchical bottleneck where errors in visual information aggregation and temporal modeling propagate to limit high-level reasoning. We further find that thinking-based reasoning is highly dependent on textual cues, improving performance with subtitles but sometimes degrading it in purely visual settings. By exposing these limitations, Video-MME-v2 establishes a demanding new testbed for the development of next-generation video MLLMs.</p> - https://arxiv.org/abs/2604.05015 - Mon, 06 Apr 2026 00:00:00 +0000 - - - The Depth Ceiling: On the Limits of Large Language Models in Discovering Latent Planning - https://arxiv.org/abs/2604.06427 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06427.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yi Xu, Philipp Jettkant, Laura Ruis</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> The viability of chain-of-thought (CoT) monitoring hinges on models being unable to reason effectively in their latent representations. Yet little is known about the limits of such latent reasoning in LLMs. We test these limits by studying whether models can discover multi-step planning strategies without supervision on intermediate steps and execute them latently, within a single forward pass. Using graph path-finding tasks that precisely control the number of required latent planning steps, we uncover a striking limitation unresolved by massive scaling: tiny transformers trained from scratch discover strategies requiring up to three latent steps, fine-tuned GPT-4o and Qwen3-32B reach five, and GPT-5.4 attains seven under few-shot prompting. Although the maximum latent planning depth models can learn during training is five, the discovered strategy generalizes up to eight latent steps at test-time. This reveals a dissociation between the ability to discover a latent strategy under final-answer supervision alone and the ability to execute it once discovered. If similar limits hold more broadly, strategies requiring multiple coordinated latent planning steps may need to be explicitly taught or externalized, lending credence to CoT monitoring.</p> - https://arxiv.org/abs/2604.06427 + TRACE: Capability-Targeted Agentic Training + https://arxiv.org/abs/2604.05336 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05336.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hangoo Kang, Tarun Suresh, Jon Saad-Falcon, Azalia Mirhoseini</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Large Language Models (LLMs) deployed in agentic environments must exercise multiple capabilities across different task instances, where a capability is performing one or more actions in a trajectory that are necessary for successfully solving a subset of tasks in the environment. Many existing approaches either rely on synthetic training data that is not targeted to the model's actual capability deficits in the target environment or train directly on the target environment, where the model needs to implicitly learn the capabilities across tasks. We introduce TRACE (Turning Recurrent Agent failures into Capability-targeted training Environments), an end-to-end system for environment-specific agent self-improvement. TRACE contrasts successful and failed trajectories to automatically identify lacking capabilities, synthesizes a targeted training environment for each that rewards whether the capability was exercised, and trains a LoRA adapter via RL on each synthetic environment, routing to the relevant adapter at inference. Empirically, TRACE generalizes across different environments, improving over the base agent by +14.1 points on τ^2-bench (customer service) and +7 perfect scores on ToolSandbox (tool use), outperforming the strongest baseline by +7.4 points and +4 perfect scores, respectively. Given the same number of rollouts, TRACE scales more efficiently than baselines, outperforming GRPO and GEPA by +9.2 and +7.4 points on τ^2-bench.</p> + https://arxiv.org/abs/2604.05336 Tue, 07 Apr 2026 00:00:00 +0000 - MMEmb-R1: Reasoning-Enhanced Multimodal Embedding with Pair-Aware Selection and Adaptive Control - https://arxiv.org/abs/2604.06156 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06156.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuchi Wang, Haiyang Yu, Weikang Bian, Jiefeng Long, Xiao Liang, Chao Feng, Hongsheng Li</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> MLLMs have been successfully applied to multimodal embedding tasks, yet their generative reasoning capabilities remain underutilized. Directly incorporating chain-of-thought reasoning into embedding learning introduces two fundamental challenges. First, structural misalignment between instance-level reasoning and pairwise contrastive supervision may lead to shortcut behavior, where the model merely learns the superficial format of reasoning. Second, reasoning is not universally beneficial for embedding tasks. Enforcing reasoning for all inputs may introduce unnecessary computation and latency, and can even obscure salient semantic signals for simple cases. To address these issues, we propose MMEmb-R1, an adaptive reasoning-based multimodal embedding framework. We formulate reasoning as a latent variable and introduce pair-aware reasoning selection that employs counterfactual intervention to identify reasoning paths beneficial for query-target alignment. Furthermore, we adopt reinforcement learning to selectively invoke reasoning only when necessary. Experiments on the MMEB-V2 benchmark demonstrate that our model achieves a score of 71.2 with only 4B parameters, establishing a new state-of-the-art while significantly reducing reasoning overhead and inference latency.</p> - https://arxiv.org/abs/2604.06156 - Tue, 07 Apr 2026 00:00:00 +0000 + Target Policy Optimization + https://arxiv.org/abs/2604.06159 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06159.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jean Kaddour</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> In RL, given a prompt, we sample a group of completions from a model and score them. Two questions follow: which completions should gain probability mass, and how should the parameters move to realize that change? Standard policy-gradient methods answer both at once, so the update can overshoot or undershoot depending on the learning rate, clipping, and other optimizer choices. We introduce Target Policy Optimization (TPO), which separates the two questions. Given scored completions, TPO constructs a target distribution q_i propto p_i^{,old} exp(u_i) and fits the policy to it by cross-entropy. The loss gradient on sampled-completion logits is p^θ- q, which vanishes once the policy matches the target. On tabular bandits, transformer sequence tasks, and billion-parameter LLM RLVR, TPO matches PG, PPO, GRPO, and DG on easy tasks and substantially outperforms them under sparse reward. Code is available at https://github.com/JeanKaddour/tpo.</p> + https://arxiv.org/abs/2604.06159 + Tue, 07 Apr 2026 17:55:59 +0000 - Graph-Based Chain-of-Thought Pruning for Reducing Redundant Reflections in Reasoning LLMs - https://arxiv.org/abs/2604.05643 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05643.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hongyuan Yuan, Xinran He, Run Shao, Bolei He, Xianwei Xue, Mengke Chen, Qiutong Pan, Haiwei Wang, Haifeng Li</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Extending CoT through RL has been widely used to enhance the reasoning capabilities of LLMs. However, due to the sparsity of reward signals, it can also induce undesirable thinking patterns such as overthinking, i.e., generating redundant intermediate reasoning content. In this work, we argue that a major source of such redundancy is inefficient reflection, which often manifests in two problematic patterns: Indiscriminate Reflection, where the model performs broad, low-impact checks throughout reasoning, and Repetitive Reflection, where it repeatedly re-verifies an already established conclusion. To address this, we introduce a graph-based CoT optimization framework. Specifically, we convert each linear CoT into a directed acyclic graph (DAG) with explicit dependency edges, and design a dual pruning strategy: branch-level pruning removes weakly contributing reflection branches, while depth-level pruning eliminates late-stage re-verification. We distill this behavior via a three-stage pipeline: (1) SFT to initialize the policy on pruned concise traces, (2) DPO to prefer correct but less redundant trajectories, and (3) GRPO with length penalty to jointly optimize answer correctness and efficiency. Experiments show that our approach reduces the average reasoning tokens by 42\% while maintaining or improving accuracy.</p> - https://arxiv.org/abs/2604.05643 - Tue, 07 Apr 2026 00:00:00 +0000 + RefineAnything: Multimodal Region-Specific Refinement for Perfect Local Details + https://arxiv.org/abs/2604.06870 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06870.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dewei Zhou, You Li, Zongxin Yang, Yi Yang</p><p><b>Upvotes:</b> 40</p><p><b>Summary:</b> We introduce region-specific image refinement as a dedicated problem setting: given an input image and a user-specified region (e.g., a scribble mask or a bounding box), the goal is to restore fine-grained details while keeping all non-edited pixels strictly unchanged. Despite rapid progress in image generation, modern models still frequently suffer from local detail collapse (e.g., distorted text, logos, and thin structures). Existing instruction-driven editing models emphasize coarse-grained semantic edits and often either overlook subtle local defects or inadvertently change the background, especially when the region of interest occupies only a small portion of a fixed-resolution input. We present RefineAnything, a multimodal diffusion-based refinement model that supports both reference-based and reference-free refinement. Building on a counter-intuitive observation that crop-and-resize can substantially improve local reconstruction under a fixed VAE input resolution, we propose Focus-and-Refine, a region-focused refinement-and-paste-back strategy that improves refinement effectiveness and efficiency by reallocating the resolution budget to the target region, while a blended-mask paste-back guarantees strict background preservation. We further introduce a boundary-aware Boundary Consistency Loss to reduce seam artifacts and improve paste-back naturalness. To support this new setting, we construct Refine-30K (20K reference-based and 10K reference-free samples) and introduce RefineEval, a benchmark that evaluates both edited-region fidelity and background consistency. On RefineEval, RefineAnything achieves strong improvements over competitive baselines and near-perfect background preservation, establishing a practical solution for high-precision local refinement. Project Page: https://limuloo.github.io/RefineAnything/.</p> + https://arxiv.org/abs/2604.06870 + Wed, 08 Apr 2026 00:00:00 +0000 - Action Images: End-to-End Policy Learning via Multiview Video Generation - https://arxiv.org/abs/2604.06168 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06168.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoyu Zhen, Zixian Gao, Qiao Sun, Yilin Zhao, Yuncong Yang, Yilun Du, Tsun-Hsuan Wang, Yi-Ling Qiao, Chuang Gan</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> World action models (WAMs) have emerged as a promising direction for robot policy learning, as they can leverage powerful video backbones to model the future states. However, existing approaches often rely on separate action modules, or use action representations that are not pixel-grounded, making it difficult to fully exploit the pretrained knowledge of video models and limiting transfer across viewpoints and environments. In this work, we present Action Images, a unified world action model that formulates policy learning as multiview video generation. Instead of encoding control as low-dimensional tokens, we translate 7-DoF robot actions into interpretable action images: multi-view action videos that are grounded in 2D pixels and explicitly track robot-arm motion. This pixel-grounded action representation allows the video backbone itself to act as a zero-shot policy, without a separate policy head or action module. Beyond control, the same unified model supports video-action joint generation, action-conditioned video generation, and action labeling under a shared representation. On RLBench and real-world evaluations, our model achieves the strongest zero-shot success rates and improves video-action joint generation quality over prior video-space world models, suggesting that interpretable action images are a promising route to policy learning.</p> - https://arxiv.org/abs/2604.06168 - Tue, 07 Apr 2026 00:00:00 +0000 - - - Qualixar OS: A Universal Operating System for AI Agent Orchestration - https://arxiv.org/abs/2604.06392 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06392.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Varun Pratap Bhardwaj</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> We present Qualixar OS, the first application-layer operating system for universal AI agent orchestration. Unlike kernel-level approaches (AIOS) or single-framework tools (AutoGen, CrewAI), Qualixar OS provides a complete runtime for heterogeneous multi-agent systems spanning 10 LLM providers, 8+ agent frameworks, and 7 transports. We contribute: (1) execution semantics for 12 multi-agent topologies including grid, forest, mesh, and maker patterns; (2) Forge, an LLM-driven team design engine with historical strategy memory; (3) three-layer model routing combining Q-learning, five strategies, and Bayesian POMDP with dynamic multi-provider discovery; (4) a consensus-based judge pipeline with Goodhart detection, JSD drift monitoring, and alignment trilemma navigation; (5) four-layer content attribution with HMAC signing and steganographic watermarks; (6) universal compatibility via the Claw Bridge supporting MCP and A2A protocols with a 25-command Universal Command Protocol; (7) a 24-tab production dashboard with visual workflow builder and skill marketplace. Qualixar OS is validated by 2,821 test cases across 217 event types and 8 quality modules. On a custom 20-task evaluation suite, the system achieves 100% accuracy at a mean cost of $0.000039 per task. Source-available under the Elastic License 2.0.</p> - https://arxiv.org/abs/2604.06392 - Tue, 07 Apr 2026 00:00:00 +0000 - - - Experience Transfer for Multimodal LLM Agents in Minecraft Game - https://arxiv.org/abs/2604.05533 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05533.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenghao Li, Jun Liu, Songbo Zhang, Huadong Jian, Hao Ni, Lik-Hang Lee, Sung-Ho Bae, Guoqing Wang, Yang Yang, Chaoning Zhang</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Multimodal LLM agents operating in complex game environments must continually reuse past experience to solve new tasks efficiently. In this work, we propose Echo, a transfer-oriented memory framework that enables agents to derive actionable knowledge from prior interactions rather than treating memory as a passive repository of static records. To make transfer explicit, Echo decomposes reusable knowledge into five dimensions: structure, attribute, process, function, and interaction. This formulation allows the agent to identify recurring patterns shared across different tasks and infer what prior experience remains applicable in new situations. Building on this formulation, Echo leverages In-Context Analogy Learning (ICAL) to retrieve relevant experiences and adapt them to unseen tasks through contextual examples. Experiments in Minecraft show that, under a from-scratch learning setting, Echo achieves a 1.3x to 1.7x speed-up on object-unlocking tasks. Moreover, Echo exhibits a burst-like chain-unlocking phenomenon, rapidly unlocking multiple similar items within a short time interval after acquiring transferable experience. These results suggest that experience transfer is a promising direction for improving the efficiency and adaptability of multimodal LLM agents in complex interactive environments.</p> - https://arxiv.org/abs/2604.05533 - Tue, 07 Apr 2026 00:00:00 +0000 - - - Graph of Skills: Dependency-Aware Structural Retrieval for Massive Agent Skills - https://arxiv.org/abs/2604.05333 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05333.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dawei Li, Zongxia Li, Hongyang Du, Xiyang Wu, Shihang Gui, Yongbei Kuang, Lichao Sun</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Skill usage has become a core component of modern agent systems and can substantially improve agents' ability to complete complex tasks. In real-world settings, where agents must monitor and interact with numerous personal applications, web browsers, and other environment interfaces, skill libraries can scale to thousands of reusable skills. Scaling to larger skill sets introduces two key challenges. First, loading the full skill set saturates the context window, driving up token costs, hallucination, and latency. In this paper, we present Graph of Skills (GoS), an inference-time structural retrieval layer for large skill libraries. GoS constructs an executable skill graph offline from skill packages, then at inference time retrieves a bounded, dependency-aware skill bundle through hybrid semantic-lexical seeding, reverse-weighted Personalized PageRank, and context-budgeted hydration. On SkillsBench and ALFWorld, GoS improves average reward by 43.6% over the vanilla full skill-loading baseline while reducing input tokens by 37.8%, and generalizes across three model families: Claude Sonnet, GPT-5.2 Codex, and MiniMax. Additional ablation studies across skill libraries ranging from 200 to 2,000 skills further demonstrate that GoS consistently outperforms both vanilla skills loading and simple vector retrieval in balancing reward, token efficiency, and runtime.</p> - https://arxiv.org/abs/2604.05333 - Tue, 07 Apr 2026 00:00:00 +0000 - - - Neural Computers - https://arxiv.org/abs/2604.06425 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06425.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingchen Zhuge, Changsheng Zhao, Haozhe Liu, Zijian Zhou, Shuming Liu, Wenyi Wang, Ernie Chang, Gael Le Lan, Junjie Fei, Wenxuan Zhang, Yasheng Sun, Zhipeng Cai, Zechun Liu, Yunyang Xiong, Yining Yang, Yuandong Tian, Yangyang Shi, Vikas Chandra, Jürgen Schmidhuber</p><p><b>Upvotes:</b> 23</p><p><b>Summary:</b> We propose a new frontier: Neural Computers (NCs) -- an emerging machine form that unifies computation, memory, and I/O in a learned runtime state. Unlike conventional computers, which execute explicit programs, agents, which act over external execution environments, and world models, which learn environment dynamics, NCs aim to make the model itself the running computer. Our long-term goal is the Completely Neural Computer (CNC): the mature, general-purpose realization of this emerging machine form, with stable execution, explicit reprogramming, and durable capability reuse. As an initial step, we study whether early NC primitives can be learned solely from collected I/O traces, without instrumented program state. Concretely, we instantiate NCs as video models that roll out screen frames from instructions, pixels, and user actions (when available) in CLI and GUI settings. These implementations show that learned runtimes can acquire early interface primitives, especially I/O alignment and short-horizon control, while routine reuse, controlled updates, and symbolic stability remain open. We outline a roadmap toward CNCs around these challenges. If overcome, CNCs could establish a new computing paradigm beyond today's agents, world models, and conventional computers.</p> - https://arxiv.org/abs/2604.06425 - Tue, 07 Apr 2026 00:00:00 +0000 - - - In-Place Test-Time Training - https://arxiv.org/abs/2604.06169 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06169.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Guhao Feng, Shengjie Luo, Kai Hua, Ge Zhang, Di He, Wenhao Huang, Tianle Cai</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> The static ``train then deploy" paradigm fundamentally limits Large Language Models (LLMs) from dynamically adapting their weights in response to continuous streams of new information inherent in real-world tasks. Test-Time Training (TTT) offers a compelling alternative by updating a subset of model parameters (fast weights) at inference time, yet its potential in the current LLM ecosystem is hindered by critical barriers including architectural incompatibility, computational inefficiency and misaligned fast weight objectives for language modeling. In this work, we introduce In-Place Test-Time Training (In-Place TTT), a framework that seamlessly endows LLMs with Test-Time Training ability. In-Place TTT treats the final projection matrix of the ubiquitous MLP blocks as its adaptable fast weights, enabling a ``drop-in" enhancement for LLMs without costly retraining from scratch. Furthermore, we replace TTT's generic reconstruction objective with a tailored, theoretically-grounded objective explicitly aligned with the Next-Token-Prediction task governing autoregressive language modeling. This principled objective, combined with an efficient chunk-wise update mechanism, results in a highly scalable algorithm compatible with context parallelism. Extensive experiments validate our framework's effectiveness: as an in-place enhancement, it enables a 4B-parameter model to achieve superior performance on tasks with contexts up to 128k, and when pretrained from scratch, it consistently outperforms competitive TTT-related approaches. Ablation study results further provide deeper insights on our design choices. Collectively, our results establish In-Place TTT as a promising step towards a paradigm of continual learning in LLMs.</p> - https://arxiv.org/abs/2604.06169 - Tue, 07 Apr 2026 00:00:00 +0000 - - - Paper Circle: An Open-source Multi-agent Research Discovery and Analysis Framework - https://arxiv.org/abs/2604.06170 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06170.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Komal Kumar, Aman Chadha, Salman Khan, Fahad Shahbaz Khan, Hisham Cholakkal</p><p><b>Upvotes:</b> 26</p><p><b>Summary:</b> The rapid growth of scientific literature has made it increasingly difficult for researchers to efficiently discover, evaluate, and synthesize relevant work. Recent advances in multi-agent large language models (LLMs) have demonstrated strong potential for understanding user intent and are being trained to utilize various tools. In this paper, we introduce Paper Circle, a multi-agent research discovery and analysis system designed to reduce the effort required to find, assess, organize, and understand academic literature. The system comprises two complementary pipelines: (1) a Discovery Pipeline that integrates offline and online retrieval from multiple sources, multi-criteria scoring, diversity-aware ranking, and structured outputs; and (2) an Analysis Pipeline that transforms individual papers into structured knowledge graphs with typed nodes such as concepts, methods, experiments, and figures, enabling graph-aware question answering and coverage verification. Both pipelines are implemented within a coder LLM-based multi-agent orchestration framework and produce fully reproducible, synchronized outputs including JSON, CSV, BibTeX, Markdown, and HTML at each agent step. This paper describes the system architecture, agent roles, retrieval and scoring methods, knowledge graph schema, and evaluation interfaces that together form the Paper Circle research workflow. We benchmark Paper Circle on both paper retrieval and paper review generation, reporting hit rate, MRR, and Recall at K. Results show consistent improvements with stronger agent models. We have publicly released the website at https://papercircle.vercel.app/ and the code at https://github.com/MAXNORM8650/papercircle.</p> - https://arxiv.org/abs/2604.06170 - Tue, 07 Apr 2026 00:00:00 +0000 - - - Beyond Accuracy: Unveiling Inefficiency Patterns in Tool-Integrated Reasoning - https://arxiv.org/abs/2604.05404 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05404.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qisheng Su, Shiting Huang, Zhen Fang, Ziyan Chen, Zehui Chen, Feng Zhao</p><p><b>Upvotes:</b> 41</p><p><b>Summary:</b> In real-world Tool-Integrated Reasoning (TIR) scenarios, where LLMs interleave reasoning with external tool calls, a major source of inefficiency is that the toolcalls create pauses between LLM requests and cause KV-Cache eviction, forcing recomputation. Also, the long, unfiltered response returned by external tools inflates the KV-Cache, so each decode step spends more time loading the growing cache and thus becomes steadily slower as context length increases. However, existing efficiency metrics like token counts and toolcall counts fail to capture the real model inference latency. To address this, we introduce PTE (Prefill Token Equivalents), a hardware-aware TIR-efficiency metric that unifies internal reasoning and external tool-use costs while explicitly accounting for non-reusable KV-Cache and long-tool-response scenarios. Validation in a high-concurrency industrial setting indicates that PTE aligns significantly better with wall-clock latency than standard token counts, while maintaining consistent efficiency rankings across diverse hardware profiles. We conduct extensive experiments across five TIR benchmarks, quantify their PTE costs, and identify four inefficiency patterns that appear in TIR. We also discover that trajectories with higher PTE costs tend to have lower reasoning correctness, indicating that simply using more tools does not improve the quality of the answer.</p> - https://arxiv.org/abs/2604.05404 - Tue, 07 Apr 2026 00:00:00 +0000 - - - RAGEN-2: Reasoning Collapse in Agentic RL - https://arxiv.org/abs/2604.06268 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06268.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zihan Wang, Chi Gui, Xing Jin, Qineng Wang, Licheng Liu, Kangrui Wang, Shiqi Chen, Linjie Li, Zhengyuan Yang, Pingyue Zhang, Yiping Lu, Jiajun Wu, Li Fei-Fei, Lijuan Wang, Yejin Choi, Manling Li</p><p><b>Upvotes:</b> 57</p><p><b>Summary:</b> RL training of multi-turn LLM agents is inherently unstable, and reasoning quality directly determines task performance. Entropy is widely used to track reasoning stability. However, entropy only measures diversity within the same input, and cannot tell whether reasoning actually responds to different inputs. In RAGEN-2, we find that even with stable entropy, models can rely on fixed templates that look diverse but are input-agnostic. We call this template collapse, a failure mode invisible to entropy and all existing metrics. To diagnose this failure, we decompose reasoning quality into within-input diversity (Entropy) and cross-input distinguishability (Mutual Information, MI), and introduce a family of mutual information proxies for online diagnosis. Across diverse tasks, mutual information correlates with final performance much more strongly than entropy, making it a more reliable proxy for reasoning quality. We further explain template collapse with a signal-to-noise ratio (SNR) mechanism. Low reward variance weakens task gradients, letting regularization terms dominate and erase cross-input reasoning differences. To address this, we propose SNR-Aware Filtering to select high-signal prompts per iteration using reward variance as a lightweight proxy. Across planning, math reasoning, web navigation, and code execution, the method consistently improves both input dependence and task performance.</p> - https://arxiv.org/abs/2604.06268 - Tue, 07 Apr 2026 00:00:00 +0000 - - - Claw-Eval: Toward Trustworthy Evaluation of Autonomous Agents - https://arxiv.org/abs/2604.06132 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06132.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bowen Ye, Rang Li, Qibin Yang, Yuanxin Liu, Linli Yao, Hanglong Lv, Zhihui Xie, Chenxin An, Lei Li, Lingpeng Kong, Qi Liu, Zhifang Sui, Tong Yang</p><p><b>Upvotes:</b> 111</p><p><b>Summary:</b> Large language models are increasingly deployed as autonomous agents executing multi-step workflows in real-world software environments. However, existing agent benchmarks suffer from three critical limitations: (1) trajectory-opaque grading that checks only final outputs, (2) underspecified safety and robustness evaluation, and (3) narrow modality coverage and interaction paradigms. We introduce Claw-Eval, an end-to-end evaluation suite addressing all three gaps. It comprises 300 human-verified tasks spanning 9 categories across three groups (general service orchestration, multimodal perception and generation, and multi-turn professional dialogue). Every agent action is recorded through three independent evidence channels (execution traces, audit logs, and environment snapshots), enabling trajectory-aware grading over 2,159 fine-grained rubric items. The scoring protocol evaluates Completion, Safety, and Robustness, reporting Average Score, Pass@k, and Pass^k across three trials to distinguish genuine capability from lucky outcomes. Experiments on 14 frontier models reveal that: (1) trajectory-opaque evaluation is systematically unreliable, missing 44% of safety violations and 13% of robustness failures that our hybrid pipeline catches; (2) controlled error injection primarily degrades consistency rather than peak capability, with Pass^3 dropping up to 24% while Pass@3 remains stable; (3) multimodal performance varies sharply, with most models performing poorer on video than on document or image, and no single model dominating across all modalities. Beyond benchmarking, Claw-Eval highlights actionable directions for agent development, shedding light on what it takes to build agents that are not only capable but reliably deployable.</p> - https://arxiv.org/abs/2604.06132 - Tue, 07 Apr 2026 00:00:00 +0000 - - - FlowInOne:Unifying Multimodal Generation as Image-in, Image-out Flow Matching - https://arxiv.org/abs/2604.06757 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06757.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Junchao Yi, Rui Zhao, Jiahao Tang, Weixian Lei, Linjie Li, Qisheng Su, Zhengyuan Yang, Lijuan Wang, Xiaofeng Zhu, Alex Jinpeng Wang</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Multimodal generation has long been dominated by text-driven pipelines where language dictates vision but cannot reason or create within it. We challenge this paradigm by asking whether all modalities, including textual descriptions, spatial layouts, and editing instructions, can be unified into a single visual representation. We present FlowInOne, a framework that reformulates multimodal generation as a purely visual flow, converting all inputs into visual prompts and enabling a clean image-in, image-out pipeline governed by a single flow matching model. This vision-centric formulation naturally eliminates cross-modal alignment bottlenecks, noise scheduling, and task-specific architectural branches, unifying text-to-image generation, layout-guided editing, and visual instruction following under one coherent paradigm. To support this, we introduce VisPrompt-5M, a large-scale dataset of 5 million visual prompt pairs spanning diverse tasks including physics-aware force dynamics and trajectory prediction, alongside VP-Bench, a rigorously curated benchmark assessing instruction faithfulness, spatial precision, visual realism, and content consistency. Extensive experiments demonstrate that FlowInOne achieves state-of-the-art performance across all unified generation tasks, surpassing both open-source models and competitive commercial systems, establishing a new foundation for fully vision-centric generative modeling where perception and creation coexist within a single continuous visual space.</p> - https://arxiv.org/abs/2604.06757 + FORGE:Fine-grained Multimodal Evaluation for Manufacturing Scenarios + https://arxiv.org/abs/2604.07413 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07413.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiangru Jian, Hao Xu, Wei Pang, Xinjian Zhao, Chengyu Tao, Qixin Zhang, Xikun Zhang, Chao Zhang, Guanzhi Deng, Alex Xue, Juan Du, Tianshu Yu, Garth Tarr, Linqi Song, Qiuzhuang Sun, Dacheng Tao</p><p><b>Upvotes:</b> 93</p><p><b>Summary:</b> The manufacturing sector is increasingly adopting Multimodal Large Language Models (MLLMs) to transition from simple perception to autonomous execution, yet current evaluations fail to reflect the rigorous demands of real-world manufacturing environments. Progress is hindered by data scarcity and a lack of fine-grained domain semantics in existing datasets. To bridge this gap, we introduce FORGE. Wefirst construct a high-quality multimodal dataset that combines real-world 2D images and 3D point clouds, annotated with fine-grained domain semantics (e.g., exact model numbers). We then evaluate 18 state-of-the-art MLLMs across three manufacturing tasks, namely workpiece verification, structural surface inspection, and assembly verification, revealing significant performance gaps. Counter to conventional understanding, the bottleneck analysis shows that visual grounding is not the primary limiting factor. Instead, insufficient domain-specific knowledge is the key bottleneck, setting a clear direction for future research. Beyond evaluation, we show that our structured annotations can serve as an actionable training resource: supervised fine-tuning of a compact 3B-parameter model on our data yields up to 90.8% relative improvement in accuracy on held-out manufacturing scenarios, providing preliminary evidence for a practical pathway toward domain-adapted manufacturing MLLMs. The code and datasets are available at https://ai4manufacturing.github.io/forge-web.</p> + https://arxiv.org/abs/2604.07413 Wed, 08 Apr 2026 00:00:00 +0000 GameWorld: Towards Standardized and Verifiable Evaluation of Multimodal Game Agents https://arxiv.org/abs/2604.07429 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07429.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingyu Ouyang, Siyuan Hu, Kevin Qinghong Lin, Hwee Tou Ng, Mike Zheng Shou</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Towards an embodied generalist for real-world interaction, Multimodal Large Language Model (MLLM) agents still suffer from challenging latency, sparse feedback, and irreversible mistakes. Video games offer an ideal testbed with rich visual observations and closed-loop interaction, demanding fine-grained perception, long-horizon planning, and precise control. However, systematically evaluating these capabilities is currently hindered by heterogeneous action interfaces and heuristic verification. To this end, we introduce GameWorld, a benchmark designed for standardized and verifiable evaluation of MLLMs as generalist game agents in browser environments. Two game agent interfaces are studied: (i) computer-use agents that directly emit keyboard and mouse controls, and (ii) generalist multimodal agents that act in a semantic action space via deterministic Semantic Action Parsing. GameWorld contains 34 diverse games and 170 tasks, each paired with state-verifiable metrics for outcome-based evaluation. The results across 18 model-interface pairs suggest that even the best performing agent is far from achieving human capabilities on video games. Extensive experiments of repeated full-benchmark reruns demonstrate the robustness of the benchmark, while further studies on real-time interaction, context-memory sensitivity, and action validity expose more challenges ahead for game agents. Together, by offering a standardized, verifiable, and reproducible evaluation framework, GameWorld lays a robust foundation for advancing research on multimodal game agents and beyond. The project page is at https://gameworld-bench.github.io.</p> + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07429.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingyu Ouyang, Siyuan Hu, Kevin Qinghong Lin, Hwee Tou Ng, Mike Zheng Shou</p><p><b>Upvotes:</b> 110</p><p><b>Summary:</b> Towards an embodied generalist for real-world interaction, Multimodal Large Language Model (MLLM) agents still suffer from challenging latency, sparse feedback, and irreversible mistakes. Video games offer an ideal testbed with rich visual observations and closed-loop interaction, demanding fine-grained perception, long-horizon planning, and precise control. However, systematically evaluating these capabilities is currently hindered by heterogeneous action interfaces and heuristic verification. To this end, we introduce GameWorld, a benchmark designed for standardized and verifiable evaluation of MLLMs as generalist game agents in browser environments. Two game agent interfaces are studied: (i) computer-use agents that directly emit keyboard and mouse controls, and (ii) generalist multimodal agents that act in a semantic action space via deterministic Semantic Action Parsing. GameWorld contains 34 diverse games and 170 tasks, each paired with state-verifiable metrics for outcome-based evaluation. The results across 18 model-interface pairs suggest that even the best performing agent is far from achieving human capabilities on video games. Extensive experiments of repeated full-benchmark reruns demonstrate the robustness of the benchmark, while further studies on real-time interaction, context-memory sensitivity, and action validity expose more challenges ahead for game agents. Together, by offering a standardized, verifiable, and reproducible evaluation framework, GameWorld lays a robust foundation for advancing research on multimodal game agents and beyond. The project page is at https://gameworld-bench.github.io.</p> https://arxiv.org/abs/2604.07429 Wed, 08 Apr 2026 00:00:00 +0000 - Personalized RewardBench: Evaluating Reward Models with Human Aligned Personalization - https://arxiv.org/abs/2604.07343 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07343.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qiyao Ma, Dechen Gao, Rui Cai, Boqi Zhao, Hanchu Zhou, Junshan Zhang, Zhe Zhao</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> Pluralistic alignment has emerged as a critical frontier in the development of Large Language Models (LLMs), with reward models (RMs) serving as a central mechanism for capturing diverse human values. While benchmarks for general response quality are prevalent, evaluating how well reward models account for individual user preferences remains an open challenge. To bridge this gap, we introduce Personalized RewardBench, a novel benchmark designed to rigorously assess reward models' capacity to model personalized preferences. We construct chosen and rejected response pairs based on strict adherence to (or violation of) user-specific rubrics, ensuring that preference distinctions are uniquely tailored to the individual. In particular, human evaluations confirm that the primary discriminative factor between pairs is strictly personal preference, with both responses maintaining high general quality (e.g., correctness, relevance and helpfulness). Extensive testing reveals that existing state-of-the-art reward models struggle significantly with personalization, peaking at an accuracy of just 75.94%. Crucially, because an effective reward model benchmark should predict a reward model's performance on downstream tasks, we conduct experiments demonstrating that our benchmark exhibits a significantly higher correlation with downstream performance in both Best-of-N (BoN) sampling and Proximal Policy Optimization (PPO) compared to existing baselines. These findings establish Personalized RewardBench as a robust and accurate proxy for evaluating reward models' performance in downstream applications.</p> - https://arxiv.org/abs/2604.07343 - Wed, 08 Apr 2026 00:00:00 +0000 - - - Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference - https://arxiv.org/abs/2604.07394 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07394.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Quantong Qiu, Zhiyi Hong, Yi Yang, Haitian Wang, Kebin Liu, Qingqing Dang, Juntao Li, Min Zhang</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> The quadratic computational complexity of standard attention mechanisms presents a severe scalability bottleneck for LLMs in long-context scenarios. While hybrid attention mechanisms combining Full Attention (FA) and Sparse Attention (SA) offer a potential solution, existing methods typically rely on static allocation ratios that fail to accommodate the variable retrieval demands of different tasks. Furthermore, head-level dynamic sparsity often introduces severe computational load imbalance and synchronization long-tails, which hinder hardware acceleration during autoregressive decoding. To bridge this gap, we introduce Flux Attention, a context-aware framework that dynamically optimizes attention computation at the layer level. By integrating a lightweight Layer Router into frozen pretrained LLMs, the proposed method adaptively routes each layer to FA or SA based on the input context. This layer-wise routing preserves high-fidelity information retrieval while ensuring contiguous memory access, translating theoretical computational reductions into practical wall-clock speedups. As a parameter-efficient approach, our framework requires only 12 hours of training on 8timesA800 GPUs. Extensive experiments across multiple long-context and mathematical reasoning benchmarks demonstrate that Flux Attention achieves a superior trade-off between performance and inference speed compared with baseline models, with speed improvements of up to 2.8times and 2.0times in the prefill and decode stages.</p> - https://arxiv.org/abs/2604.07394 - Wed, 08 Apr 2026 00:00:00 +0000 - - - TC-AE: Unlocking Token Capacity for Deep Compression Autoencoders - https://arxiv.org/abs/2604.07340 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07340.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Teng Li, Ziyuan Huang, Cong Chen, Yangfu Li, Yuanhuiyi Lyu, Dandan Zheng, Chunhua Shen, Jun Zhang</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> We propose TC-AE, a ViT-based architecture for deep compression autoencoders. Existing methods commonly increase the channel number of latent representations to maintain reconstruction quality under high compression ratios. However, this strategy often leads to latent representation collapse, which degrades generative performance. Instead of relying on increasingly complex architectures or multi-stage training schemes, TC-AE addresses this challenge from the perspective of the token space, the key bridge between pixels and image latents, through two complementary innovations: Firstly, we study token number scaling by adjusting the patch size in ViT under a fixed latent budget, and identify aggressive token-to-latent compression as the key factor that limits effective scaling. To address this issue, we decompose token-to-latent compression into two stages, reducing structural information loss and enabling effective token number scaling for generation. Secondly, to further mitigate latent representation collapse, we enhance the semantic structure of image tokens via joint self-supervised training, leading to more generative-friendly latents. With these designs, TC-AE achieves substantially improved reconstruction and generative performance under deep compression. We hope our research will advance ViT-based tokenizer for visual generation.</p> - https://arxiv.org/abs/2604.07340 - Wed, 08 Apr 2026 00:00:00 +0000 - - - FP4 Explore, BF16 Train: Diffusion Reinforcement Learning via Efficient Rollout Scaling - https://arxiv.org/abs/2604.06916 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06916.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yitong Li, Junsong Chen, Shuchen Xue, Pengcuo Zeren, Siyuan Fu, Dinghao Yang, Yangyang Tang, Junjie Bai, Ping Luo, Song Han, Enze Xie</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> Reinforcement-Learning-based post-training has recently emerged as a promising paradigm for aligning text-to-image diffusion models with human preferences. In recent studies, increasing the rollout group size yields pronounced performance improvements, indicating substantial room for further alignment gains. However, scaling rollouts on large-scale foundational diffusion models (e.g., FLUX.1-12B) imposes a heavy computational burden. To alleviate this bottleneck, we explore the integration of FP4 quantization into Diffusion RL rollouts. Yet, we identify that naive quantized pipelines inherently introduce risks of performance degradation. To overcome this dilemma between efficiency and training integrity, we propose Sol-RL (Speed-of-light RL), a novel FP4-empowered Two-stage Reinforcement Learning framework. First, we utilize high-throughput NVFP4 rollouts to generate a massive candidate pool and extract a highly contrastive subset. Second, we regenerate these selected samples in BF16 precision and optimize the policy exclusively on them. By decoupling candidate exploration from policy optimization, Sol-RL integrates the algorithmic mechanisms of rollout scaling with the system-level throughput gains of NVFP4. This synergistic algorithm-hardware design effectively accelerates the rollout phase while reserving high-fidelity samples for optimization. We empirically demonstrate that our framework maintains the training integrity of BF16 precision pipeline while fully exploiting the throughput gains enabled by FP4 arithmetic. Extensive experiments across SANA, FLUX.1, and SD3.5-L substantiate that our approach delivers superior alignment performance across multiple metrics while accelerating training convergence by up to 4.64times, unlocking the power of massive rollout scaling at a fraction of the cost.</p> - https://arxiv.org/abs/2604.06916 - Wed, 08 Apr 2026 00:00:00 +0000 - - - MARS: Enabling Autoregressive Models Multi-Token Generation - https://arxiv.org/abs/2604.07023 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07023.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ziqi Jin, Lei Wang, Ziwei Luo, Aixin Sun</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> Autoregressive (AR) language models generate text one token at a time, even when consecutive tokens are highly predictable given earlier context. We introduce MARS (Mask AutoRegreSsion), a lightweight fine-tuning method that teaches an instruction-tuned AR model to predict multiple tokens per forward pass. MARS adds no architectural modifications, no extra parameters, and produces a single model that can still be called exactly like the original AR model with no performance degradation. Unlike speculative decoding, which maintains a separate draft model alongside the target, or multi-head approaches such as Medusa, which attach additional prediction heads, MARS requires only continued training on existing instruction data. When generating one token per forward pass, MARS matches or exceeds the AR baseline on six standard benchmarks. When allowed to accept multiple tokens per step, it maintains baseline-level accuracy while achieving 1.5-1.7x throughput. We further develop a block-level KV caching strategy for batch inference, achieving up to 1.71x wall-clock speedup over AR with KV cache on Qwen2.5-7B. Finally, MARS supports real-time speed adjustment via confidence thresholding: under high request load, the serving system can increase throughput on the fly without swapping models or restarting, providing a practical latency-quality knob for deployment.</p> - https://arxiv.org/abs/2604.07023 - Wed, 08 Apr 2026 00:00:00 +0000 - - - INSPATIO-WORLD: A Real-Time 4D World Simulator via Spatiotemporal Autoregressive Modeling - https://arxiv.org/abs/2604.07209 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07209.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> InSpatio Team, Donghui Shen, Guofeng Zhang, Haomin Liu, Haoyu Ji, Hujun Bao, Hongjia Zhai, Jialin Liu, Jing Guo, Nan Wang, Siji Pan, Weihong Pan, Weijian Xie, Xianbin Liu, Xiaojun Xiang, Xiaoyu Zhang, Xinyu Chen, Yifu Wang, Yipeng Chen, Zhenzhou Fan, Zhewen Le, Zhichao Ye, Ziqiang Zhao</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> Building world models with spatial consistency and real-time interactivity remains a fundamental challenge in computer vision. Current video generation paradigms often struggle with a lack of spatial persistence and insufficient visual realism, making it difficult to support seamless navigation in complex environments. To address these challenges, we propose INSPATIO-WORLD, a novel real-time framework capable of recovering and generating high-fidelity, dynamic interactive scenes from a single reference video. At the core of our approach is a Spatiotemporal Autoregressive (STAR) architecture, which enables consistent and controllable scene evolution through two tightly coupled components: Implicit Spatiotemporal Cache aggregates reference and historical observations into a latent world representation, ensuring global consistency during long-horizon navigation; Explicit Spatial Constraint Module enforces geometric structure and translates user interactions into precise and physically plausible camera trajectories. Furthermore, we introduce Joint Distribution Matching Distillation (JDMD). By using real-world data distributions as a regularizing guide, JDMD effectively overcomes the fidelity degradation typically caused by over-reliance on synthetic data. Extensive experiments demonstrate that INSPATIO-WORLD significantly outperforms existing state-of-the-art (SOTA) models in spatial consistency and interaction precision, ranking first among real-time interactive methods on the WorldScore-Dynamic benchmark, and establishing a practical pipeline for navigating 4D environments reconstructed from monocular videos.</p> - https://arxiv.org/abs/2604.07209 - Wed, 08 Apr 2026 00:00:00 +0000 - - - OpenSpatial: A Principled Data Engine for Empowering Spatial Intelligence - https://arxiv.org/abs/2604.07296 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07296.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jianhui Liu, Haoze Sun, Wenbo Li, Yanbing Zhang, Rui Yang, Zhiliang Zhu, Yijun Yang, Shenghe Zheng, Nan Jiang, Jiaxiu Jiang, Haoyang Huang, Tien-Tsin Wong, Nan Duan, Xiaojuan Qi</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> Spatial understanding is a fundamental cornerstone of human-level intelligence. Nonetheless, current research predominantly focuses on domain-specific data production, leaving a critical void: the absence of a principled, open-source engine capable of fully unleashing the potential of high-quality spatial data. To bridge this gap, we elucidate the design principles of a robust data generation system and introduce OpenSpatial -- an open-source data engine engineered for high quality, extensive scalability, broad task diversity, and optimized efficiency. OpenSpatial adopts 3D bounding boxes as the fundamental primitive to construct a comprehensive data hierarchy across five foundational tasks: Spatial Measurement (SM), Spatial Relationship (SR), Camera Perception (CP), Multi-view Consistency (MC), and Scene-Aware Reasoning (SAR). Leveraging this scalable infrastructure, we curate OpenSpatial-3M, a large-scale dataset comprising 3 million high-fidelity samples. Extensive evaluations demonstrate that versatile models trained on our dataset achieve state-of-the-art performance across a wide spectrum of spatial reasoning benchmarks. Notably, the best-performing model exhibits a substantial average improvement of 19 percent, relatively. Furthermore, we provide a systematic analysis of how data attributes influence spatial perception. By open-sourcing both the engine and the 3M-scale dataset, we provide a robust foundation to accelerate future research in spatial intelligence.</p> - https://arxiv.org/abs/2604.07296 - Wed, 08 Apr 2026 00:00:00 +0000 - - - Think in Strokes, Not Pixels: Process-Driven Image Generation via Interleaved Reasoning - https://arxiv.org/abs/2604.04746 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.04746.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Lei Zhang, Junjiao Tian, Zhipeng Fan, Kunpeng Li, Jialiang Wang, Weifeng Chen, Markos Georgopoulos, Felix Juefei-Xu, Yuxiang Bao, Julian McAuley, Manling Li, Zecheng He</p><p><b>Upvotes:</b> 64</p><p><b>Summary:</b> Humans paint images incrementally: they plan a global layout, sketch a coarse draft, inspect, and refine details, and most importantly, each step is grounded in the evolving visual states. However, can unified multimodal models trained on text-image interleaved datasets also imagine the chain of intermediate states? In this paper, we introduce process-driven image generation, a multi-step paradigm that decomposes synthesis into an interleaved reasoning trajectory of thoughts and actions. Rather than generating images in a single step, our approach unfolds across multiple iterations, each consisting of 4 stages: textual planning, visual drafting, textual reflection, and visual refinement. The textual reasoning explicitly conditions how the visual state should evolve, while the generated visual intermediate in turn constrains and grounds the next round of textual reasoning. A core challenge of process-driven generation stems from the ambiguity of intermediate states: how can models evaluate each partially-complete image? We address this through dense, step-wise supervision that maintains two complementary constraints: for the visual intermediate states, we enforce the spatial and semantic consistency; for the textual intermediate states, we preserve the prior visual knowledge while enabling the model to identify and correct prompt-violating elements. This makes the generation process explicit, interpretable, and directly supervisable. To validate proposed method, we conduct experiments under various text-to-image generation benchmarks.</p> - https://arxiv.org/abs/2604.04746 - Wed, 08 Apr 2026 00:00:00 +0000 - - - HY-Embodied-0.5: Embodied Foundation Models for Real-World Agents - https://arxiv.org/abs/2604.07430 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07430.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tencent Robotics X, HY Vision Team, Xumin Yu, Zuyan Liu, Ziyi Wang, He Zhang, Yongming Rao, Fangfu Liu, Yani Zhang, Ruowen Zhao, Oran Wang, Yves Liang, Haitao Lin, Minghui Wang, Yubo Dong, Kevin Cheng, Bolin Ni, Rui Huang, Han Hu, Zhengyou Zhang, Linus, Shunyu Yao</p><p><b>Upvotes:</b> 152</p><p><b>Summary:</b> We introduce HY-Embodied-0.5, a family of foundation models specifically designed for real-world embodied agents. To bridge the gap between general Vision-Language Models (VLMs) and the demands of embodied agents, our models are developed to enhance the core capabilities required by embodied intelligence: spatial and temporal visual perception, alongside advanced embodied reasoning for prediction, interaction, and planning. The HY-Embodied-0.5 suite comprises two primary variants: an efficient model with 2B activated parameters designed for edge deployment, and a powerful model with 32B activated parameters targeted for complex reasoning. To support the fine-grained visual perception essential for embodied tasks, we adopt a Mixture-of-Transformers (MoT) architecture to enable modality-specific computing. By incorporating latent tokens, this design effectively enhances the perceptual representation of the models. To improve reasoning capabilities, we introduce an iterative, self-evolving post-training paradigm. Furthermore, we employ on-policy distillation to transfer the advanced capabilities of the large model to the smaller variant, thereby maximizing the performance potential of the compact model. Extensive evaluations across 22 benchmarks, spanning visual perception, spatial reasoning, and embodied understanding, demonstrate the effectiveness of our approach. Our MoT-2B model outperforms similarly sized state-of-the-art models on 16 benchmarks, while the 32B variant achieves performance comparable to frontier models such as Gemini 3.0 Pro. In downstream robot control experiments, we leverage our robust VLM foundation to train an effective Vision-Language-Action (VLA) model, achieving compelling results in real-world physical evaluations. Code and models are open-sourced at https://github.com/Tencent-Hunyuan/HY-Embodied.</p> - https://arxiv.org/abs/2604.07430 - Wed, 08 Apr 2026 00:00:00 +0000 - - - Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability - https://arxiv.org/abs/2604.06628 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.06628.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qihan Ren, Peng Wang, Ruikun Cai, Shuai Shao, Dadi Guo, Yuejin Xie, Yafu Li, Quanshi Zhang, Xia Hu, Jing Shao, Dongrui Liu</p><p><b>Upvotes:</b> 292</p><p><b>Summary:</b> A prevailing narrative in LLM post-training holds that supervised finetuning (SFT) memorizes while reinforcement learning (RL) generalizes. We revisit this claim for reasoning SFT with long chain-of-thought (CoT) supervision and find that cross-domain generalization is not absent but conditional, jointly shaped by optimization dynamics, training data, and base-model capability. Some reported failures are under-optimization artifacts: cross-domain performance first degrades before recovering and improving with extended training (a dip-and-recovery pattern), so shorttraining checkpoints can underestimate generalization. Data quality and structure both matter: low-quality solutions broadly hurt generalization,while verified long-CoT traces yield consistent cross-domain gains. Model capability is essential: stronger models internalize transferable procedural patterns (e.g., backtracking) even from a toy arithmetic game, while weaker ones imitate surface verbosity. This generalization is asymmetric, however: reasoning improves while safety degrades, reframing the question from whether reasoning SFT generalizes to under what conditions and at what cost.</p> - https://arxiv.org/abs/2604.06628 - Wed, 08 Apr 2026 00:00:00 +0000 - - - Towards Real-world Human Behavior Simulation: Benchmarking Large Language Models on Long-horizon, Cross-scenario, Heterogeneous Behavior Traces - https://arxiv.org/abs/2604.08362 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08362.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiawei Chen, Ruoxi Xu, Boxi Cao, Ruotong Pan, Yunfei Zhang, Yifei Hu, Yong Du, Tingting Gao, Yaojie Lu, Yingfei Sun, Xianpei Han, Le Sun, Xiangyu Wu, Hongyu Lin</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> The emergence of Large Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing benchmarks remain constrained to isolated scenarios, narrow action spaces, or synthetic data, failing to capture the holistic nature of authentic human behavior. To bridge this gap, we introduce OmniBehavior, the first user simulation benchmark constructed entirely from real-world data, integrating long-horizon, cross-scenario, and heterogeneous behavioral patterns into a unified framework. Based on this benchmark, we first provide empirical evidence that previous datasets with isolated scenarios suffer from tunnel vision, whereas real-world decision-making relies on long-term, cross-scenario causal chains. Extensive evaluations of state-of-the-art LLMs reveal that current models struggle to accurately simulate these complex behaviors, with performance plateauing even as context windows expand. Crucially, a systematic comparison between simulated and authentic behaviors uncovers a fundamental structural bias: LLMs tend to converge toward a positive average person, exhibiting hyper-activity, persona homogenization, and a Utopian bias. This results in the loss of individual differences and long-tail behaviors, highlighting critical directions for future high-fidelity simulation research.</p> - https://arxiv.org/abs/2604.08362 + p1: Better Prompt Optimization with Fewer Prompts + https://arxiv.org/abs/2604.08801 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08801.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhaolin Gao, Yu, Wang, Bo Liu, Thorsten Joachims, Kianté Brantley, Wen Sun</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Prompt optimization improves language models without updating their weights by searching for a better system prompt, but its effectiveness varies widely across tasks. We study what makes a task amenable to prompt optimization. We show that the reward variance across different system prompts can be decomposed into two components: variance among responses, which captures generation stochasticity, and variance among system prompts, which captures differences in system prompt quality. Prompt optimization succeeds when variance among system prompts is sufficiently large, but fails when variance among responses dominates the variance of the system prompts. Surprisingly, we further show that scaling to more user prompts can hurt optimization by reducing variance among system prompts, especially on heterogeneous datasets where different user prompts favor different system prompts. Motivated by this insight, we propose p1, a simple user prompt filtering method that selects a small subset of user prompts with high variance across candidate system prompts. This subset of user prompts allows one to distinguish a good system prompt from a bad one, making system optimization easier. Experiments on reasoning benchmarks show that p1 substantially improves prompt optimization over training on the full dataset and outperforms strong baselines such as GEPA. Notably, training on only two prompts from AIME 24 yields a system prompt that generalizes well to other reasoning benchmarks.</p> + https://arxiv.org/abs/2604.08801 Thu, 09 Apr 2026 00:00:00 +0000 - SIM1: Physics-Aligned Simulator as Zero-Shot Data Scaler in Deformable Worlds - https://arxiv.org/abs/2604.08544 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08544.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yunsong Zhou, Hangxu Liu, Xuekun Jiang, Xing Shen, Yuanzhen Zhou, Hui Wang, Baole Fang, Yang Tian, Mulin Yu, Qiaojun Yu, Li Ma, Hengjie Li, Hanqing Wang, Jia Zeng, Jiangmiao Pang</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Robotic manipulation with deformable objects represents a data-intensive regime in embodied learning, where shape, contact, and topology co-evolve in ways that far exceed the variability of rigids. Although simulation promises relief from the cost of real-world data acquisition, prevailing sim-to-real pipelines remain rooted in rigid-body abstractions, producing mismatched geometry, fragile soft dynamics, and motion primitives poorly suited for cloth interaction. We posit that simulation fails not for being synthetic, but for being ungrounded. To address this, we introduce SIM1, a physics-aligned real-to-sim-to-real data engine that grounds simulation in the physical world. Given limited demonstrations, the system digitizes scenes into metric-consistent twins, calibrates deformable dynamics through elastic modeling, and expands behaviors via diffusion-based trajectory generation with quality filtering. This pipeline transforms sparse observations into scaled synthetic supervision with near-demonstration fidelity. Experiments show that policies trained on purely synthetic data achieve parity with real-data baselines at a 1:15 equivalence ratio, while delivering 90% zero-shot success and 50% generalization gains in real-world deployment. These results validate physics-aligned simulation as scalable supervision for deformable manipulation and a practical pathway for data-efficient policy learning.</p> - https://arxiv.org/abs/2604.08544 + ReconPhys: Reconstruct Appearance and Physical Attributes from Single Video + https://arxiv.org/abs/2604.07882 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07882.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Boyuan Wang, Xiaofeng Wang, Yongkang Li, Zheng Zhu, Yifan Chang, Angen Ye, Guosheng Zhao, Chaojun Ni, Guan Huang, Yijie Ren, Yueqi Duan, Xingang Wang</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Reconstructing non-rigid objects with physical plausibility remains a significant challenge. Existing approaches leverage differentiable rendering for per-scene optimization, recovering geometry and dynamics but requiring expensive tuning or manual annotation, which limits practicality and generalizability. To address this, we propose ReconPhys, the first feedforward framework that jointly learns physical attribute estimation and 3D Gaussian Splatting reconstruction from a single monocular video. Our method employs a dual-branch architecture trained via a self-supervised strategy, eliminating the need for ground-truth physics labels. Given a video sequence, ReconPhys simultaneously infers geometry, appearance, and physical attributes. Experiments on a large-scale synthetic dataset demonstrate superior performance: our method achieves 21.64 PSNR in future prediction compared to 13.27 by state-of-the-art optimization baselines, while reducing Chamfer Distance from 0.349 to 0.004. Crucially, ReconPhys enables fast inference (<1 second) versus hours required by existing methods, facilitating rapid generation of simulation-ready assets for robotics and graphics.</p> + https://arxiv.org/abs/2604.07882 Thu, 09 Apr 2026 00:00:00 +0000 - ViVa: A Video-Generative Value Model for Robot Reinforcement Learning - https://arxiv.org/abs/2604.08168 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08168.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jindi Lv, Hao Li, Jie Li, Yifei Nie, Fankun Kong, Yang Wang, Xiaofeng Wang, Zheng Zhu, Chaojun Ni, Qiuping Deng, Hengtao Li, Jiancheng Lv, Guan Huang</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Vision-language-action (VLA) models have advanced robot manipulation through large-scale pretraining, but real-world deployment remains challenging due to partial observability and delayed feedback. Reinforcement learning addresses this via value functions, which assess task progress and guide policy improvement. However, existing value models built on vision-language models (VLMs) struggle to capture temporal dynamics, undermining reliable value estimation in long-horizon tasks. In this paper, we propose ViVa, a video-generative value model that repurposes a pretrained video generator for value estimation. Taking the current observation and robot proprioception as input, ViVa jointly predicts future proprioception and a scalar value for the current state. By leveraging the spatiotemporal priors of a pretrained video generator, our approach grounds value estimation in anticipated embodiment dynamics, moving beyond static snapshots to intrinsically couple value with foresight. Integrated into RECAP, ViVa delivers substantial improvements on real-world box assembly. Qualitative analysis across all three tasks confirms that ViVa produces more reliable value signals, accurately reflecting task progress. By leveraging spatiotemporal priors from video corpora, ViVa also generalizes to novel objects, highlighting the promise of video-generative models for value estimation.</p> - https://arxiv.org/abs/2604.08168 + Efficient RL Training for LLMs with Experience Replay + https://arxiv.org/abs/2604.08706 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08706.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Charles Arnal, Vivien Cabannes, Taco Cohen, Julia Kempe, Remi Munos</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> While Experience Replay - the practice of storing rollouts and reusing them multiple times during training - is a foundational technique in general RL, it remains largely unexplored in LLM post-training due to the prevailing belief that fresh, on-policy data is essential for high performance. In this work, we challenge this assumption. We present a systematic study of replay buffers for LLM post-training, formalizing the optimal design as a trade-off between staleness-induced variance, sample diversity and the high computational cost of generation. We show that strict on-policy sampling is suboptimal when generation is expensive. Empirically, we show that a well-designed replay buffer can drastically reduce inference compute without degrading - and in some cases even improving - final model performance, while preserving policy entropy.</p> + https://arxiv.org/abs/2604.08706 Thu, 09 Apr 2026 00:00:00 +0000 - Small Vision-Language Models are Smart Compressors for Long Video Understanding - https://arxiv.org/abs/2604.08120 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08120.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Junjie Fei, Jun Chen, Zechun Liu, Yunyang Xiong, Chong Zhou, Wei Wen, Junlin Han, Mingchen Zhuge, Saksham Suri, Qi Qian, Shuming Liu, Lemeng Wu, Raghuraman Krishnamoorthi, Vikas Chandra, Mohamed Elhoseiny, Chenchen Zhu</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Adapting Multimodal Large Language Models (MLLMs) for hour-long videos is bottlenecked by context limits. Dense visual streams saturate token budgets and exacerbate the lost-in-the-middle phenomenon. Existing heuristics, like sparse sampling or uniform pooling, blindly sacrifice fidelity by discarding decisive moments and wasting bandwidth on irrelevant backgrounds. We propose Tempo, an efficient query-aware framework compressing long videos for downstream understanding. Tempo leverages a Small Vision-Language Model (SVLM) as a local temporal compressor, casting token reduction as an early cross-modal distillation process to generate compact, intent-aligned representations in a single forward pass. To enforce strict budgets without breaking causality, we introduce Adaptive Token Allocation (ATA). Exploiting the SVLM's zero-shot relevance prior and semantic front-loading, ATA acts as a training-free O(1) dynamic router. It allocates dense bandwidth to query-critical segments while compressing redundancies into minimal temporal anchors to maintain the global storyline. Extensive experiments show our 6B architecture achieves state-of-the-art performance with aggressive dynamic compression (0.5-16 tokens/frame). On the extreme-long LVBench (4101s), Tempo scores 52.3 under a strict 8K visual budget, outperforming GPT-4o and Gemini 1.5 Pro. Scaling to 2048 frames reaches 53.7. Crucially, Tempo compresses hour-long videos substantially below theoretical limits, proving true long-form video understanding relies on intent-driven efficiency rather than greedily padded context windows.</p> - https://arxiv.org/abs/2604.08120 + Uni-ViGU: Towards Unified Video Generation and Understanding via A Diffusion-Based Video Generator + https://arxiv.org/abs/2604.08121 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08121.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Luozheng Qin, Jia Gong, Qian Qiao, Tianjiao Li, Li Xu, Haoyu Pan, Chao Qu, Zhiyu Tan, Hao Li</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> Unified multimodal models integrating visual understanding and generation face a fundamental challenge: visual generation incurs substantially higher computational costs than understanding, particularly for video. This imbalance motivates us to invert the conventional paradigm: rather than extending understanding-centric MLLMs to support generation, we propose Uni-ViGU, a framework that unifies video generation and understanding by extending a video generator as the foundation. We introduce a unified flow method that performs continuous flow matching for video and discrete flow matching for text within a single process, enabling coherent multimodal generation. We further propose a modality-driven MoE-based framework that augments Transformer blocks with lightweight layers for text generation while preserving generative priors. To repurpose generation knowledge for understanding, we design a bidirectional training mechanism with two stages: Knowledge Recall reconstructs input prompts to leverage learned text-video correspondences, while Capability Refinement fine-tunes on detailed captions to establish discriminative shared representations. Experiments demonstrate that Uni-ViGU achieves competitive performance on both video generation and understanding, validating generation-centric architectures as a scalable path toward unified multimodal intelligence. Project Page and Code: https://fr0zencrane.github.io/uni-vigu-page/.</p> + https://arxiv.org/abs/2604.08121 Thu, 09 Apr 2026 00:00:00 +0000 - Structured Distillation of Web Agent Capabilities Enables Generalization - https://arxiv.org/abs/2604.07776 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07776.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xing Han Lù, Siva Reddy</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Frontier LLMs can navigate complex websites, but their cost and reliance on third-party APIs make local deployment impractical. We introduce Agent-as-Annotators, a framework that structures synthetic trajectory generation for web agents by analogy to human annotation roles, replacing the Task Designer, Annotator, and Supervisor with modular LLM components. Using Gemini 3 Pro as teacher, we generate 3,000 trajectories across six web environments and fine-tune a 9B-parameter student with pure supervised learning on the 2,322 that pass quality filtering. The resulting model achieves 41.5% on WebArena, surpassing closed-source models such as Claude 3.5 Sonnet (36.0%) and GPT-4o (31.5%) under the same evaluation protocol, and nearly doubling the previous best open-weight result (Go-Browse, 21.7%). Capabilities transfer to unseen environments, with an 18.2 percentage point gain on WorkArena L1 (an enterprise platform never seen during training) and consistent improvements across three additional benchmarks. Ablations confirm that each pipeline component contributes meaningfully, with Judge filtering, evaluation hints, and reasoning traces each accounting for measurable gains. These results demonstrate that structured trajectory synthesis from a single frontier teacher is sufficient to produce competitive, locally deployable web agents. Project page: https://agent-as-annotators.github.io</p> - https://arxiv.org/abs/2604.07776 + EXAONE 4.5 Technical Report + https://arxiv.org/abs/2604.08644 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08644.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Eunbi Choi, Kibong Choi, Sehyun Chun, Seokhee Hong, Junwon Hwang, Hyojin Jeon, Ahra Jo, Hyunjik Jo, Yeonsik Jo, Joonkee Kim, Seonghwan Kim, Soyeon Kim, Sunkyoung Kim, Yireun Kim, Yongil Kim, Changhun Lee, Haeju Lee, Jinsik Lee, Kyungmin Lee, Sangha Park, Kwangrok Ryoo, Minju Seo, Sejong Yang, Heuiyeen Yeen, Hwan Chang, Stanley Jungkyu Choi, Yejin Choi, Kyubeen Han, Joonwon Jang, Kijeong Jeon, Geunyeong Jeong, Gerrard Jeongwon Jo, Jiyeon Jung, Daeseong Kim, Dohoon Kim, Dohyun Kim, Hyunseo Kim, Minu Kim, Myoungshin Kim, Youchul Kim, Byungoh Ko, Christopher Lee, Edward Hwayoung Lee, Honglak Lee, Jiyoung Lee, Sangeun Lee, Seungwon Lim, Woohyung Lim, Jueun Mun, Jaewoo Park, Jimin Park, Jinho Park, Yongmin Park, Wooseok Seo, Yongwoo Song, Sihyuk Yi, Kyungjae Yoo, Sangyeon Yoon</p><p><b>Upvotes:</b> 62</p><p><b>Summary:</b> This technical report introduces EXAONE 4.5, the first open-weight vision language model released by LG AI Research. EXAONE 4.5 is architected by integrating a dedicated visual encoder into the existing EXAONE 4.0 framework, enabling native multimodal pretraining over both visual and textual modalities. The model is trained on large-scale data with careful curation, particularly emphasizing document-centric corpora that align with LG's strategic application domains. This targeted data design enables substantial performance gains in document understanding and related tasks, while also delivering broad improvements across general language capabilities. EXAONE 4.5 extends context length up to 256K tokens, facilitating long-context reasoning and enterprise-scale use cases. Comparative evaluations demonstrate that EXAONE 4.5 achieves competitive performance in general benchmarks while outperforming state-of-the-art models of similar scale in document understanding and Korean contextual reasoning. As part of LG's ongoing effort toward practical industrial deployment, EXAONE 4.5 is designed to be continuously extended with additional domains and application scenarios to advance AI for a better life.</p> + https://arxiv.org/abs/2604.08644 Thu, 09 Apr 2026 00:00:00 +0000 - FIT: A Large-Scale Dataset for Fit-Aware Virtual Try-On - https://arxiv.org/abs/2604.08526 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08526.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Johanna Karras, Yuanhao Wang, Yingwei Li, Ira Kemelmacher-Shlizerman</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Given a person and a garment image, virtual try-on (VTO) aims to synthesize a realistic image of the person wearing the garment, while preserving their original pose and identity. Although recent VTO methods excel at visualizing garment appearance, they largely overlook a crucial aspect of the try-on experience: the accuracy of garment fit -- for example, depicting how an extra-large shirt looks on an extra-small person. A key obstacle is the absence of datasets that provide precise garment and body size information, particularly for "ill-fit" cases, where garments are significantly too large or too small. Consequently, current VTO methods default to generating well-fitted results regardless of the garment or person size. In this paper, we take the first steps towards solving this open problem. We introduce FIT (Fit-Inclusive Try-on), a large-scale VTO dataset comprising over 1.13M try-on image triplets accompanied by precise body and garment measurements. We overcome the challenges of data collection via a scalable synthetic strategy: (1) We programmatically generate 3D garments using GarmentCode and drape them via physics simulation to capture realistic garment fit. (2) We employ a novel re-texturing framework to transform synthetic renderings into photorealistic images while strictly preserving geometry. (3) We introduce person identity preservation into our re-texturing model to generate paired person images (same person, different garments) for supervised training. Finally, we leverage our FIT dataset to train a baseline fit-aware virtual try-on model. Our data and results set the new state-of-the-art for fit-aware virtual try-on, as well as offer a robust benchmark for future research. We will make all data and code publicly available on our project page: https://johannakarras.github.io/FIT.</p> - https://arxiv.org/abs/2604.08526 + WildDet3D: Scaling Promptable 3D Detection in the Wild + https://arxiv.org/abs/2604.08626 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08626.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Weikai Huang, Jieyu Zhang, Sijun Li, Taoyang Jia, Jiafei Duan, Yunqian Cheng, Jaemin Cho, Mattew Wallingford, Rustin Soraki, Chris Dongjoo Kim, Donovan Clay, Taira Anderson, Winson Han, Ali Farhadi, Bharath Hariharan, Zhongzheng Ren, Ranjay Krishna</p><p><b>Upvotes:</b> 238</p><p><b>Summary:</b> Understanding objects in 3D from a single image is a cornerstone of spatial intelligence. A key step toward this goal is monocular 3D object detection--recovering the extent, location, and orientation of objects from an input RGB image. To be practical in the open world, such a detector must generalize beyond closed-set categories, support diverse prompt modalities, and leverage geometric cues when available. Progress is hampered by two bottlenecks: existing methods are designed for a single prompt type and lack a mechanism to incorporate additional geometric cues, and current 3D datasets cover only narrow categories in controlled environments, limiting open-world transfer. In this work we address both gaps. First, we introduce WildDet3D, a unified geometry-aware architecture that natively accepts text, point, and box prompts and can incorporate auxiliary depth signals at inference time. Second, we present WildDet3D-Data, the largest open 3D detection dataset to date, constructed by generating candidate 3D boxes from existing 2D annotations and retaining only human-verified ones, yielding over 1M images across 13.5K categories in diverse real-world scenes. WildDet3D establishes a new state-of-the-art across multiple benchmarks and settings. In the open-world setting, it achieves 22.6/24.8 AP3D on our newly introduced WildDet3D-Bench with text and box prompts. On Omni3D, it reaches 34.2/36.4 AP3D with text and box prompts, respectively. In zero-shot evaluation, it achieves 40.3/48.9 ODS on Argoverse 2 and ScanNet. Notably, incorporating depth cues at inference time yields substantial additional gains (+20.7 AP on average across settings).</p> + https://arxiv.org/abs/2604.08626 Thu, 09 Apr 2026 00:00:00 +0000 - OmniJigsaw: Enhancing Omni-Modal Reasoning via Modality-Orchestrated Reordering - https://arxiv.org/abs/2604.08209 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08209.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yiduo Jia, Muzhi Zhu, Hao Zhong, Mingyu Liu, Yuling Xi, Hao Chen, Bin Qin, Yongjie Yang, Zhenbo Luo, Chunhua Shen</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> To extend the reinforcement learning post-training paradigm to omni-modal models for concurrently bolstering video-audio understanding and collaborative reasoning, we propose OmniJigsaw, a generic self-supervised framework built upon a temporal reordering proxy task. Centered on the chronological reconstruction of shuffled audio-visual clips, this paradigm strategically orchestrates visual and auditory signals to compel cross-modal integration through three distinct strategies: Joint Modality Integration, Sample-level Modality Selection, and Clip-level Modality Masking. Recognizing that the efficacy of such proxy tasks is fundamentally tied to puzzle quality, we design a two-stage coarse-to-fine data filtering pipeline, which facilitates the efficient adaptation of OmniJigsaw to massive unannotated omni-modal data. Our analysis reveals a ``bi-modal shortcut phenomenon'' in joint modality integration and demonstrates that fine-grained clip-level modality masking mitigates this issue while outperforming sample-level modality selection. Extensive evaluations on 15 benchmarks show substantial gains in video, audio, and collaborative reasoning, validating OmniJigsaw as a scalable paradigm for self-supervised omni-modal learning.</p> - https://arxiv.org/abs/2604.08209 - Thu, 09 Apr 2026 00:00:00 +0000 + VisionFoundry: Teaching VLMs Visual Perception with Synthetic Images + https://arxiv.org/abs/2604.09531 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09531.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Guanyu Zhou, Yida Yin, Wenhao Chai, Shengbang Tong, Xingyu Fu, Zhuang Liu</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Vision-language models (VLMs) still struggle with visual perception tasks such as spatial understanding and viewpoint recognition. One plausible contributing factor is that natural image datasets provide limited supervision for low-level visual skills. This motivates a practical question: can targeted synthetic supervision, generated from only a task keyword such as Depth Order, address these weaknesses? To investigate this question, we introduce VisionFoundry, a task-aware synthetic data generation pipeline that takes only the task name as input and uses large language models (LLMs) to generate questions, answers, and text-to-image (T2I) prompts, then synthesizes images with T2I models and verifies consistency with a proprietary VLM, requiring no reference images or human annotation. Using VisionFoundry, we construct VisionFoundry-10K, a synthetic visual question answering (VQA) dataset containing 10k image-question-answer triples spanning 10 tasks. Models trained on VisionFoundry-10K achieve substantial improvements on visual perception benchmarks: +7% on MMVP and +10% on CV-Bench-3D, while preserving broader capabilities and showing favorable scaling behavior as data size increases. Our results suggest that limited task-targeted supervision is an important contributor to this bottleneck and that synthetic supervision is a promising path toward more systematic training for VLMs.</p> + https://arxiv.org/abs/2604.09531 + Fri, 10 Apr 2026 00:00:00 +0000 - MolmoWeb: Open Visual Web Agent and Open Data for the Open Web - https://arxiv.org/abs/2604.08516 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08516.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tanmay Gupta, Piper Wolters, Zixian Ma, Peter Sushko, Rock Yuren Pang, Diego Llanes, Yue Yang, Taira Anderson, Boyuan Zheng, Zhongzheng Ren, Harsh Trivedi, Taylor Blanton, Caleb Ouellette, Winson Han, Ali Farhadi, Ranjay Krishna</p><p><b>Upvotes:</b> 35</p><p><b>Summary:</b> Web agents--autonomous systems that navigate and execute tasks on the web on behalf of users--have the potential to transform how people interact with the digital world. However, the most capable web agents today rely on proprietary models with undisclosed training data and recipes, limiting scientific understanding, reproducibility, and community-driven progress. We believe agents for the open web should be built in the open. To this end, we introduce (1) MolmoWebMix, a large and diverse mixture of browser task demonstrations and web-GUI perception data and (2) MolmoWeb, a family of fully open multimodal web agents. Specifically, MolmoWebMix combines over 100K synthetic task trajectories from multiple complementary generation pipelines with 30K+ human demonstrations, atomic web-skill trajectories, and GUI perception data, including referring expression grounding and screenshot question answering. MolmoWeb agents operate as instruction-conditioned visual-language action policies: given a task instruction and a webpage screenshot, they predict the next browser action, requiring no access to HTML, accessibility trees, or specialized APIs. Available in 4B and 8B size, on browser-use benchmarks like WebVoyager, Online-Mind2Web, and DeepShop, MolmoWeb agents achieve state-of-the-art results outperforming similar scale open-weight-only models such as Fara-7B, UI-Tars-1.5-7B, and Holo1-7B. MolmoWeb-8B also surpasses set-of-marks (SoM) agents built on much larger closed frontier models like GPT-4o. We further demonstrate consistent gains through test-time scaling via parallel rollouts with best-of-N selection, achieving 94.7% and 60.5% pass@4 (compared to 78.2% and 35.3% pass@1) on WebVoyager and Online-Mind2Web respectively. We will release model checkpoints, training data, code, and a unified evaluation harness to enable reproducibility and accelerate open research on web agents.</p> - https://arxiv.org/abs/2604.08516 - Thu, 09 Apr 2026 00:00:00 +0000 + ScheMatiQ: From Research Question to Structured Data through Interactive Schema Discovery + https://arxiv.org/abs/2604.09237 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09237.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shahar Levy, Eliya Habba, Reshef Mintz, Barak Raveh, Renana Keydar, Gabriel Stanovsky</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Many disciplines pose natural-language research questions over large document collections whose answers typically require structured evidence, traditionally obtained by manually designing an annotation schema and exhaustively labeling the corpus, a slow and error-prone process. We introduce ScheMatiQ, which leverages calls to a backbone LLM to take a question and a corpus to produce a schema and a grounded database, with a web interface that lets steer and revise the extraction. In collaboration with domain experts, we show that ScheMatiQ yields outputs that support real-world analysis in law and computational biology. We release ScheMatiQ as open source with a public web interface, and invite experts across disciplines to use it with their own data. All resources, including the website, source code, and demonstration video, are available at: www.ScheMatiQ-ai.com</p> + https://arxiv.org/abs/2604.09237 + Fri, 10 Apr 2026 00:00:00 +0000 - Act Wisely: Cultivating Meta-Cognitive Tool Use in Agentic Multimodal Models - https://arxiv.org/abs/2604.08545 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08545.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shilin Yan, Jintao Tong, Hongwei Xue, Xiaojun Tang, Yangyang Wang, Kunyu Shi, Guannan Zhang, Ruixuan Li, Yixiong Zou</p><p><b>Upvotes:</b> 37</p><p><b>Summary:</b> The advent of agentic multimodal models has empowered systems to actively interact with external environments. However, current agents suffer from a profound meta-cognitive deficit: they struggle to arbitrate between leveraging internal knowledge and querying external utilities. Consequently, they frequently fall prey to blind tool invocation, resorting to reflexive tool execution even when queries are resolvable from the raw visual context. This pathological behavior precipitates severe latency bottlenecks and injects extraneous noise that derails sound reasoning. Existing reinforcement learning protocols attempt to mitigate this via a scalarized reward that penalizes tool usage. Yet, this coupled formulation creates an irreconcilable optimization dilemma: an aggressive penalty suppresses essential tool use, whereas a mild penalty is entirely subsumed by the variance of the accuracy reward during advantage normalization, rendering it impotent against tool overuse. To transcend this bottleneck, we propose HDPO, a framework that reframes tool efficiency from a competing scalar objective to a strictly conditional one. By eschewing reward scalarization, HDPO maintains two orthogonal optimization channels: an accuracy channel that maximizes task correctness, and an efficiency channel that enforces execution economy exclusively within accurate trajectories via conditional advantage estimation. This decoupled architecture naturally induces a cognitive curriculum-compelling the agent to first master task resolution before refining its self-reliance. Extensive evaluations demonstrate that our resulting model, Metis, reduces tool invocations by orders of magnitude while simultaneously elevating reasoning accuracy.</p> - https://arxiv.org/abs/2604.08545 - Thu, 09 Apr 2026 00:00:00 +0000 + Envisioning the Future, One Step at a Time + https://arxiv.org/abs/2604.09527 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09527.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Stefan Andreas Baumann, Jannik Wiese, Tommaso Martorella, Mahdi M. Kalayeh, Björn Ommer</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Accurately anticipating how complex, diverse scenes will evolve requires models that represent uncertainty, simulate along extended interaction chains, and efficiently explore many plausible futures. Yet most existing approaches rely on dense video or latent-space prediction, expending substantial capacity on dense appearance rather than on the underlying sparse trajectories of points in the scene. This makes large-scale exploration of future hypotheses costly and limits performance when long-horizon, multi-modal motion is essential. We address this by formulating the prediction of open-set future scene dynamics as step-wise inference over sparse point trajectories. Our autoregressive diffusion model advances these trajectories through short, locally predictable transitions, explicitly modeling the growth of uncertainty over time. This dynamics-centric representation enables fast rollout of thousands of diverse futures from a single image, optionally guided by initial constraints on motion, while maintaining physical plausibility and long-range coherence. We further introduce OWM, a benchmark for open-set motion prediction based on diverse in-the-wild videos, to evaluate accuracy and variability of predicted trajectory distributions under real-world uncertainty. Our method matches or surpasses dense simulators in predictive accuracy while achieving orders-of-magnitude higher sampling speed, making open-set future prediction both scalable and practical. Project page: http://compvis.github.io/myriad.</p> + https://arxiv.org/abs/2604.09527 + Fri, 10 Apr 2026 00:00:00 +0000 - Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering - https://arxiv.org/abs/2604.08224 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08224.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenyu Zhou, Huacan Chai, Wenteng Chen, Zihan Guo, Rong Shan, Yuanyi Song, Tianyi Xu, Yingxuan Yang, Aofan Yu, Weiming Zhang, Congming Zheng, Jiachen Zhu, Zeyu Zheng, Zhuosheng Zhang, Xingyu Lou, Changwang Zhang, Zhihui Fu, Jun Wang, Weiwen Liu, Jianghao Lin, Weinan Zhang</p><p><b>Upvotes:</b> 40</p><p><b>Summary:</b> Large language model (LLM) agents are increasingly built less by changing model weights than by reorganizing the runtime around them. Capabilities that earlier systems expected the model to recover internally are now externalized into memory stores, reusable skills, interaction protocols, and the surrounding harness that makes these modules reliable in practice. This paper reviews that shift through the lens of externalization. Drawing on the idea of cognitive artifacts, we argue that agent infrastructure matters not merely because it adds auxiliary components, but because it transforms hard cognitive burdens into forms that the model can solve more reliably. Under this view, memory externalizes state across time, skills externalize procedural expertise, protocols externalize interaction structure, and harness engineering serves as the unification layer that coordinates them into governed execution. We trace a historical progression from weights to context to harness, analyze memory, skills, and protocols as three distinct but coupled forms of externalization, and examine how they interact inside a larger agent system. We further discuss the trade-off between parametric and externalized capability, identify emerging directions such as self-evolving harnesses and shared agent infrastructure, and discuss open challenges in evaluation, governance, and the long-term co-evolution of models and external infrastructure. The result is a systems-level framework for explaining why practical agent progress increasingly depends not only on stronger models, but on better external cognitive infrastructure.</p> - https://arxiv.org/abs/2604.08224 - Thu, 09 Apr 2026 00:00:00 +0000 + Many-Tier Instruction Hierarchy in LLM Agents + https://arxiv.org/abs/2604.09443 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09443.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jingyu Zhang, Tianjian Li, William Jurayj, Hongyuan Zhan, Benjamin Van Durme, Daniel Khashabi</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> Large language model agents receive instructions from many sources-system messages, user prompts, tool outputs, and more-each carrying different levels of trust and authority. When these instructions conflict, models must reliably follow the highest-privilege instruction to remain safe and effective. The dominant paradigm, instruction hierarchy (IH), assumes a fixed, small set of privilege levels (typically fewer than five) defined by rigid role labels (e.g., system > user). This is inadequate for real-world agentic settings, where conflicts can arise across far more sources and contexts. In this work, we propose Many-Tier Instruction Hierarchy (ManyIH), a paradigm for resolving instruction conflicts among instructions with arbitrarily many privilege levels. We introduce ManyIH-Bench, the first benchmark for ManyIH. ManyIH-Bench requires models to navigate up to 12 levels of conflicting instructions with varying privileges, comprising 853 agentic tasks (427 coding and 426 instruction-following). ManyIH-Bench composes constraints developed by LLMs and verified by humans to create realistic and difficult test cases spanning 46 real-world agents. Our experiments show that even the current frontier models perform poorly (~40% accuracy) when instruction conflict scales. This work underscores the urgent need for methods that explicitly target fine-grained, scalable instruction conflict resolution in agentic settings.</p> + https://arxiv.org/abs/2604.09443 + Fri, 10 Apr 2026 00:00:00 +0000 - KnowU-Bench: Towards Interactive, Proactive, and Personalized Mobile Agent Evaluation - https://arxiv.org/abs/2604.08455 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08455.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tongbo Chen, Zhengxi Lu, Zhan Xu, Guocheng Shao, Shaohan Zhao, Fei Tang, Yong Du, Kaitao Song, Yizhou Liu, Yuchen Yan, Wenqi Zhang, Xu Tan, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen</p><p><b>Upvotes:</b> 40</p><p><b>Summary:</b> Personalized mobile agents that infer user preferences and calibrate proactive assistance hold great promise as everyday digital assistants, yet existing benchmarks fail to capture what this requires. Prior work evaluates preference recovery from static histories or intent prediction from fixed contexts. Neither tests whether an agent can elicit missing preferences through interaction, nor whether it can decide when to intervene, seek consent, or remain silent in a live GUI environment. We introduce KnowU-Bench, an online benchmark for personalized mobile agents built on a reproducible Android emulation environment, covering 42 general GUI tasks, 86 personalized tasks, and 64 proactive tasks. Unlike prior work that treats user preferences as static context, KnowU-Bench hides the user profile from the agent and exposes only behavioral logs, forcing genuine preference inference rather than context lookup. To support multi-turn preference elicitation, it instantiates an LLM-driven user simulator grounded in structured profiles, enabling realistic clarification dialogues and proactive consent handling. Beyond personalization, KnowU-Bench provides comprehensive evaluation of the complete proactive decision chain, including grounded GUI execution, consent negotiation, and post-rejection restraint, evaluated through a hybrid protocol combining rule-based verification with LLM-as-a-Judge scoring. Our experiments reveal a striking degradation: agents that excel at explicit task execution fall below 50% under vague instructions requiring user preference inference or intervention calibration, even for frontier models like Claude Sonnet 4.6. The core bottlenecks are not GUI navigation but preference acquisition and intervention calibration, exposing a fundamental gap between competent interface operation and trustworthy personal assistance.</p> - https://arxiv.org/abs/2604.08455 - Thu, 09 Apr 2026 00:00:00 +0000 + Hierarchical SVG Tokenization: Learning Compact Visual Programs for Scalable Vector Graphics Modeling + https://arxiv.org/abs/2604.05072 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.05072.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ximing Xing, Ziteng Xue, Zhenxi Li, Weicong Liang, Linqing Wang, Zhantao Yang, Tiankai Hang, Zijin Yin, Qinglin Lu, Chunyu Wang, Qian Yu</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Recent large language models have shifted SVG generation from differentiable rendering optimization to autoregressive program synthesis. However, existing approaches still rely on generic byte-level tokenization inherited from natural language processing, which poorly reflects the geometric structure of vector graphics. Numerical coordinates are fragmented into discrete symbols, destroying spatial relationships and introducing severe token redundancy, often leading to coordinate hallucination and inefficient long-sequence generation. To address these challenges, we propose HiVG, a hierarchical SVG tokenization framework tailored for autoregressive vector graphics generation. HiVG decomposes raw SVG strings into structured atomic tokens and further compresses executable command--parameter groups into geometry-constrained segment tokens, substantially improving sequence efficiency while preserving syntactic validity. To further mitigate spatial mismatch, we introduce a Hierarchical Mean--Noise (HMN) initialization strategy that injects numerical ordering signals and semantic priors into new token embeddings. Combined with a curriculum training paradigm that progressively increases program complexity, HiVG enables more stable learning of executable SVG programs. Extensive experiments on both text-to-SVG and image-to-SVG tasks demonstrate improved generation fidelity, spatial consistency, and sequence efficiency compared with conventional tokenization schemes. Our code is publicly available at https://github.com/ximinng/HiVG</p> + https://arxiv.org/abs/2604.05072 + Fri, 10 Apr 2026 00:00:00 +0000 - DMax: Aggressive Parallel Decoding for dLLMs - https://arxiv.org/abs/2604.08302 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08302.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zigeng Chen, Gongfan Fang, Xinyin Ma, Ruonan Yu, Xinchao Wang</p><p><b>Upvotes:</b> 43</p><p><b>Summary:</b> We present DMax, a new paradigm for efficient diffusion language models (dLLMs). It mitigates error accumulation in parallel decoding, enabling aggressive decoding parallelism while preserving generation quality. Unlike conventional masked dLLMs that decode through a binary mask-to-token transition, DMax reformulates decoding as a progressive self-refinement from mask embeddings to token embeddings. At the core of our approach is On-Policy Uniform Training, a novel training strategy that efficiently unifies masked and uniform dLLMs, equipping the model to recover clean tokens from both masked inputs and its own erroneous predictions. Building on this foundation, we further propose Soft Parallel Decoding. We represent each intermediate decoding state as an interpolation between the predicted token embedding and the mask embedding, enabling iterative self-revising in embedding space. Extensive experiments across a variety of benchmarks demonstrate the effectiveness of DMax. Compared with the original LLaDA-2.0-mini, our method improves TPF on GSM8K from 2.04 to 5.47 while preserving accuracy. On MBPP, it increases TPF from 2.71 to 5.86 while maintaining comparable performance. On two H200 GPUs, our model achieves an average of 1,338 TPS at batch size 1. Code is available at: https://github.com/czg1225/DMax</p> - https://arxiv.org/abs/2604.08302 - Thu, 09 Apr 2026 00:00:00 +0000 + ELT: Elastic Looped Transformers for Visual Generation + https://arxiv.org/abs/2604.09168 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09168.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sahil Goyal, Swayam Agrawal, Gautham Govind Anil, Prateek Jain, Sujoy Paul, Aditya Kusupati</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> We introduce Elastic Looped Transformers (ELT), a highly parameter-efficient class of visual generative models based on a recurrent transformer architecture. While conventional generative models rely on deep stacks of unique transformer layers, our approach employs iterative, weight-shared transformer blocks to drastically reduce parameter counts while maintaining high synthesis quality. To effectively train these models for image and video generation, we propose the idea of Intra-Loop Self Distillation (ILSD), where student configurations (intermediate loops) are distilled from the teacher configuration (maximum training loops) to ensure consistency across the model's depth in a single training step. Our framework yields a family of elastic models from a single training run, enabling Any-Time inference capability with dynamic trade-offs between computational cost and generation quality, with the same parameter count. ELT significantly shifts the efficiency frontier for visual synthesis. With 4times reduction in parameter count under iso-inference-compute settings, ELT achieves a competitive FID of 2.0 on class-conditional ImageNet 256 times 256 and FVD of 72.8 on class-conditional UCF-101.</p> + https://arxiv.org/abs/2604.09168 + Fri, 10 Apr 2026 00:00:00 +0000 - OpenVLThinkerV2: A Generalist Multimodal Reasoning Model for Multi-domain Visual Tasks - https://arxiv.org/abs/2604.08539 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08539.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wenbo Hu, Xin Chen, Yan Gao-Tian, Yihe Deng, Nanyun Peng, Kai-Wei Chang</p><p><b>Upvotes:</b> 44</p><p><b>Summary:</b> Group Relative Policy Optimization (GRPO) has emerged as the de facto Reinforcement Learning (RL) objective driving recent advancements in Multimodal Large Language Models. However, extending this success to open-source multimodal generalist models remains heavily constrained by two primary challenges: the extreme variance in reward topologies across diverse visual tasks, and the inherent difficulty of balancing fine-grained perception with multi-step reasoning capabilities. To address these issues, we introduce Gaussian GRPO (G^2RPO), a novel RL training objective that replaces standard linear scaling with non-linear distributional matching. By mathematically forcing the advantage distribution of any given task to strictly converge to a standard normal distribution, N(0,1), G^2RPO theoretically ensures inter-task gradient equity, mitigates vulnerabilities to heavy-tail outliers, and offers symmetric update for positive and negative rewards. Leveraging the enhanced training stability provided by G^2RPO, we introduce two task-level shaping mechanisms to seamlessly balance perception and reasoning. First, response length shaping dynamically elicits extended reasoning chains for complex queries while enforce direct outputs to bolster visual grounding. Second, entropy shaping tightly bounds the model's exploration zone, effectively preventing both entropy collapse and entropy explosion. Integrating these methodologies, we present OpenVLThinkerV2, a highly robust, general-purpose multimodal model. Extensive evaluations across 18 diverse benchmarks demonstrate its superior performance over strong open-source and leading proprietary frontier models.</p> - https://arxiv.org/abs/2604.08539 - Thu, 09 Apr 2026 00:00:00 +0000 + ECHO: Efficient Chest X-ray Report Generation with One-step Block Diffusion + https://arxiv.org/abs/2604.09450 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09450.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Lifeng Chen, Tianqi You, Hao Liu, Zhimin Bao, Jile Jiao, Xiao Han, Zhicai Ou, Tao Sun, Xiaofeng Mou, Xiaojie Jin, Yi Xu</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> Chest X-ray report generation (CXR-RG) has the potential to substantially alleviate radiologists' workload. However, conventional autoregressive vision--language models (VLMs) suffer from high inference latency due to sequential token decoding. Diffusion-based models offer a promising alternative through parallel generation, but they still require multiple denoising iterations. Compressing multi-step denoising to a single step could further reduce latency, but often degrades textual coherence due to the mean-field bias introduced by token-factorized denoisers. To address this challenge, we propose ECHO, an efficient diffusion-based VLM (dVLM) for chest X-ray report generation. ECHO enables stable one-step-per-block inference via a novel Direct Conditional Distillation (DCD) framework, which mitigates the mean-field limitation by constructing unfactorized supervision from on-policy diffusion trajectories to encode joint token dependencies. In addition, we introduce a Response-Asymmetric Diffusion (RAD) training strategy that further improves training efficiency while maintaining model effectiveness. Extensive experiments demonstrate that ECHO surpasses state-of-the-art autoregressive methods, improving RaTE and SemScore by 64.33\% and 60.58\% respectively, while achieving an 8times inference speedup without compromising clinical accuracy.</p> + https://arxiv.org/abs/2604.09450 + Fri, 10 Apr 2026 00:00:00 +0000 - LPM 1.0: Video-based Character Performance Model - https://arxiv.org/abs/2604.07823 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.07823.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ailing Zeng, Casper Yang, Chauncey Ge, Eddie Zhang, Garvey Xu, Gavin Lin, Gilbert Gu, Jeremy Pi, Leo Li, Mingyi Shi, Sheng Bi, Steven Tang, Thorn Hang, Tobey Guo, Vincent Li, Xin Tong, Yikang Li, Yuchen Sun, Yue, Zhao, Yuhan Lu, Yuwei Li, Zane Zhang, Zeshi Yang, Zi Ye</p><p><b>Upvotes:</b> 55</p><p><b>Summary:</b> Performance, the externalization of intent, emotion, and personality through visual, vocal, and temporal behavior, is what makes a character alive. Learning such performance from video is a promising alternative to traditional 3D pipelines. However, existing video models struggle to jointly achieve high expressiveness, real-time inference, and long-horizon identity stability, a tension we call the performance trilemma. Conversation is the most comprehensive performance scenario, as characters simultaneously speak, listen, react, and emote while maintaining identity over time. To address this, we present LPM 1.0 (Large Performance Model), focusing on single-person full-duplex audio-visual conversational performance. Concretely, we build a multimodal human-centric dataset through strict filtering, speaking-listening audio-video pairing, performance understanding, and identity-aware multi-reference extraction; train a 17B-parameter Diffusion Transformer (Base LPM) for highly controllable, identity-consistent performance through multimodal conditioning; and distill it into a causal streaming generator (Online LPM) for low-latency, infinite-length interaction. At inference, given a character image with identity-aware references, LPM 1.0 generates listening videos from user audio and speaking videos from synthesized audio, with text prompts for motion control, all at real-time speed with identity-stable, infinite-length generation. LPM 1.0 thus serves as a visual engine for conversational agents, live streaming characters, and game NPCs. To systematically evaluate this setting, we propose LPM-Bench, the first benchmark for interactive character performance. LPM 1.0 achieves state-of-the-art results across all evaluated dimensions while maintaining real-time inference.</p> - https://arxiv.org/abs/2604.07823 - Thu, 09 Apr 2026 00:00:00 +0000 + BERT-as-a-Judge: A Robust Alternative to Lexical Methods for Efficient Reference-Based LLM Evaluation + https://arxiv.org/abs/2604.09497 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09497.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hippolyte Gisserot-Boukhlef, Nicolas Boizard, Emmanuel Malherbe, Céline Hudelot, Pierre Colombo</p><p><b>Upvotes:</b> 29</p><p><b>Summary:</b> Accurate evaluation is central to the large language model (LLM) ecosystem, guiding model selection and downstream adoption across diverse use cases. In practice, however, evaluating generative outputs typically relies on rigid lexical methods to extract and assess answers, which can conflate a model's true problem-solving ability with its compliance with predefined formatting guidelines. While recent LLM-as-a-Judge approaches mitigate this issue by assessing semantic correctness rather than strict structural conformity, they also introduce substantial computational overhead, making evaluation costly. In this work, we first systematically investigate the limitations of lexical evaluation through a large-scale empirical study spanning 36 models and 15 downstream tasks, demonstrating that such methods correlate poorly with human judgments. To address this limitation, we introduce BERT-as-a-Judge, an encoder-driven approach for assessing answer correctness in reference-based generative settings, robust to variations in output phrasing, and requiring only lightweight training on synthetically annotated question-candidate-reference triplets. We show that it consistently outperforms the lexical baseline while matching the performance of much larger LLM judges, providing a compelling tradeoff between the two and enabling reliable, scalable evaluation. Finally, through extensive experimentation, we provide detailed insights into BERT-as-a-Judge's performance to offer practical guidance for practitioners, and release all project artifacts to foster downstream adoption.</p> + https://arxiv.org/abs/2604.09497 + Fri, 10 Apr 2026 00:00:00 +0000 - MegaStyle: Constructing Diverse and Scalable Style Dataset via Consistent Text-to-Image Style Mapping - https://arxiv.org/abs/2604.08364 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08364.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Junyao Gao, Sibo Liu, Jiaxing Li, Yanan Sun, Yuanpeng Tu, Fei Shen, Weidong Zhang, Cairong Zhao, Jun Zhang</p><p><b>Upvotes:</b> 92</p><p><b>Summary:</b> In this paper, we introduce MegaStyle, a novel and scalable data curation pipeline that constructs an intra-style consistent, inter-style diverse and high-quality style dataset. We achieve this by leveraging the consistent text-to-image style mapping capability of current large generative models, which can generate images in the same style from a given style description. Building on this foundation, we curate a diverse and balanced prompt gallery with 170K style prompts and 400K content prompts, and generate a large-scale style dataset MegaStyle-1.4M via content-style prompt combinations. With MegaStyle-1.4M, we propose style-supervised contrastive learning to fine-tune a style encoder MegaStyle-Encoder for extracting expressive, style-specific representations, and we also train a FLUX-based style transfer model MegaStyle-FLUX. Extensive experiments demonstrate the importance of maintaining intra-style consistency, inter-style diversity and high-quality for style dataset, as well as the effectiveness of the proposed MegaStyle-1.4M. Moreover, when trained on MegaStyle-1.4M, MegaStyle-Encoder and MegaStyle-FLUX provide reliable style similarity measurement and generalizable style transfer, making a significant contribution to the style transfer community. More results are available at our project website https://jeoyal.github.io/MegaStyle/.</p> - https://arxiv.org/abs/2604.08364 - Thu, 09 Apr 2026 00:00:00 +0000 + SPPO: Sequence-Level PPO for Long-Horizon Reasoning Tasks + https://arxiv.org/abs/2604.08865 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08865.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tianyi Wang, Yixia Li, Long Li, Yibiao Chen, Shaohan Huang, Yun Chen, Peng Li, Yang Liu, Guanhua Chen</p><p><b>Upvotes:</b> 29</p><p><b>Summary:</b> Proximal Policy Optimization (PPO) is central to aligning Large Language Models (LLMs) in reasoning tasks with verifiable rewards. However, standard token-level PPO struggles in this setting due to the instability of temporal credit assignment over long Chain-of-Thought (CoT) horizons and the prohibitive memory cost of the value model. While critic-free alternatives like GRPO mitigate these issues, they incur significant computational overhead by requiring multiple samples for baseline estimation, severely limiting training throughput. In this paper, we introduce Sequence-Level PPO (SPPO), a scalable algorithm that harmonizes the sample efficiency of PPO with the stability of outcome-based updates. SPPO reformulates the reasoning process as a Sequence-Level Contextual Bandit problem, employing a decoupled scalar value function to derive low-variance advantage signals without multi-sampling. Extensive experiments on mathematical benchmarks demonstrate that SPPO significantly surpasses standard PPO and matches the performance of computation-heavy group-based methods, offering a resource-efficient framework for aligning reasoning LLMs.</p> + https://arxiv.org/abs/2604.08865 + Fri, 10 Apr 2026 00:00:00 +0000 - When Numbers Speak: Aligning Textual Numerals and Visual Instances in Text-to-Video Diffusion Models - https://arxiv.org/abs/2604.08546 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08546.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhengyang Sun, Yu Chen, Xin Zhou, Xiaofan Li, Xiwu Chen, Dingkang Liang, Xiang Bai</p><p><b>Upvotes:</b> 109</p><p><b>Summary:</b> Text-to-video diffusion models have enabled open-ended video synthesis, but often struggle with generating the correct number of objects specified in a prompt. We introduce NUMINA , a training-free identify-then-guide framework for improved numerical alignment. NUMINA identifies prompt-layout inconsistencies by selecting discriminative self- and cross-attention heads to derive a countable latent layout. It then refines this layout conservatively and modulates cross-attention to guide regeneration. On the introduced CountBench, NUMINA improves counting accuracy by up to 7.4% on Wan2.1-1.3B, and by 4.9% and 5.5% on 5B and 14B models, respectively. Furthermore, CLIP alignment is improved while maintaining temporal consistency. These results demonstrate that structural guidance complements seed search and prompt enhancement, offering a practical path toward count-accurate text-to-video diffusion. The code is available at https://github.com/H-EmbodVis/NUMINA.</p> - https://arxiv.org/abs/2604.08546 - Thu, 09 Apr 2026 00:00:00 +0000 + Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory + https://arxiv.org/abs/2604.08995 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08995.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zile Wang, Zexiang Liu, Jaixing Li, Kaichen Huang, Baixin Xu, Fei Kang, Mengyin An, Peiyu Wang, Biao Jiang, Yichen Wei, Yidan Xietian, Jiangbo Pei, Liang Hu, Boyi Jiang, Hua Xue, Zidong Wang, Haofeng Sun, Wei Li, Wanli Ouyang, Xianglong He, Yang Liu, Yangguang Li, Yahui Zhou</p><p><b>Upvotes:</b> 46</p><p><b>Summary:</b> With the advancement of interactive video generation, diffusion models have increasingly demonstrated their potential as world models. However, existing approaches still struggle to simultaneously achieve memory-enabled long-term temporal consistency and high-resolution real-time generation, limiting their applicability in real-world scenarios. To address this, we present Matrix-Game 3.0, a memory-augmented interactive world model designed for 720p real-time longform video generation. Building upon Matrix-Game 2.0, we introduce systematic improvements across data, model, and inference. First, we develop an upgraded industrial-scale infinite data engine that integrates Unreal Engine-based synthetic data, large-scale automated collection from AAA games, and real-world video augmentation to produce high-quality Video-Pose-Action-Prompt quadruplet data at scale. Second, we propose a training framework for long-horizon consistency: by modeling prediction residuals and re-injecting imperfect generated frames during training, the base model learns self-correction; meanwhile, camera-aware memory retrieval and injection enable the base model to achieve long horizon spatiotemporal consistency. Third, we design a multi-segment autoregressive distillation strategy based on Distribution Matching Distillation (DMD), combined with model quantization and VAE decoder pruning, to achieve efficient real-time inference. Experimental results show that Matrix-Game 3.0 achieves up to 40 FPS real-time generation at 720p resolution with a 5B model, while maintaining stable memory consistency over minute-long sequences. Scaling up to a 2x14B model further improves generation quality, dynamics, and generalization. Our approach provides a practical pathway toward industrial-scale deployable world models.</p> + https://arxiv.org/abs/2604.08995 + Fri, 10 Apr 2026 00:00:00 +0000 - ClawBench: Can AI Agents Complete Everyday Online Tasks? - https://arxiv.org/abs/2604.08523 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08523.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuxuan Zhang, Yubo Wang, Yipeng Zhu, Penghui Du, Junwen Miao, Xuan Lu, Wendong Xu, Yunzhuo Hao, Songcheng Cai, Xiaochen Wang, Huaisong Zhang, Xian Wu, Yi Lu, Minyi Lei, Kai Zou, Huifeng Yin, Ping Nie, Liang Chen, Dongfu Jiang, Wenhu Chen, Kelsey R. Allen</p><p><b>Upvotes:</b> 242</p><p><b>Summary:</b> AI agents may be able to automate your inbox, but can they automate other routine aspects of your life? Everyday online tasks offer a realistic yet unsolved testbed for evaluating the next generation of AI agents. To this end, we introduce ClawBench, an evaluation framework of 153 simple tasks that people need to accomplish regularly in their lives and work, spanning 144 live platforms across 15 categories, from completing purchases and booking appointments to submitting job applications. These tasks require demanding capabilities beyond existing benchmarks, such as obtaining relevant information from user-provided documents, navigating multi-step workflows across diverse platforms, and write-heavy operations like filling in many detailed forms correctly. Unlike existing benchmarks that evaluate agents in offline sandboxes with static pages, ClawBench operates on production websites, preserving the full complexity, dynamic nature, and challenges of real-world web interaction. A lightweight interception layer captures and blocks only the final submission request, ensuring safe evaluation without real-world side effects. Our evaluations of 7 frontier models show that both proprietary and open-source models can complete only a small portion of these tasks. For example, Claude Sonnet 4.6 achieves only 33.3%. Progress on ClawBench brings us closer to AI agents that can function as reliable general-purpose assistants.</p> - https://arxiv.org/abs/2604.08523 - Thu, 09 Apr 2026 00:00:00 +0000 + Strips as Tokens: Artist Mesh Generation with Native UV Segmentation + https://arxiv.org/abs/2604.09132 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09132.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Rui Xu, Dafei Qin, Kaichun Qiao, Qiujie Dong, Huaijin Pi, Qixuan Zhang, Longwen Zhang, Lan Xu, Jingyi Yu, Wenping Wang, Taku Komura</p><p><b>Upvotes:</b> 50</p><p><b>Summary:</b> Recent advancements in autoregressive transformers have demonstrated remarkable potential for generating artist-quality meshes. However, the token ordering strategies employed by existing methods typically fail to meet professional artist standards, where coordinate-based sorting yields inefficiently long sequences, and patch-based heuristics disrupt the continuous edge flow and structural regularity essential for high-quality modeling. To address these limitations, we propose Strips as Tokens (SATO), a novel framework with a token ordering strategy inspired by triangle strips. By constructing the sequence as a connected chain of faces that explicitly encodes UV boundaries, our method naturally preserves the organized edge flow and semantic layout characteristic of artist-created meshes. A key advantage of this formulation is its unified representation, enabling the same token sequence to be decoded into either a triangle or quadrilateral mesh. This flexibility facilitates joint training on both data types: large-scale triangle data provides fundamental structural priors, while high-quality quad data enhances the geometric regularity of the outputs. Extensive experiments demonstrate that SATO consistently outperforms prior methods in terms of geometric quality, structural coherence, and UV segmentation.</p> + https://arxiv.org/abs/2604.09132 + Fri, 10 Apr 2026 00:00:00 +0000 - SkillClaw: Let Skills Evolve Collectively with Agentic Evolver - https://arxiv.org/abs/2604.08377 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.08377.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ziyu Ma, Shidong Yang, Yuxiang Ji, Xucong Wang, Yong Wang, Yiming Hu, Tongwen Huang, Xiangxiang Chu</p><p><b>Upvotes:</b> 258</p><p><b>Summary:</b> Large language model (LLM) agents such as OpenClaw rely on reusable skills to perform complex tasks, yet these skills remain largely static after deployment. As a result, similar workflows, tool usage patterns, and failure modes are repeatedly rediscovered across users, preventing the system from improving with experience. While interactions from different users provide complementary signals about when a skill works or fails, existing systems lack a mechanism to convert such heterogeneous experiences into reliable skill updates. To address these issues, we present SkillClaw, a framework for collective skill evolution in multi-user agent ecosystems, which treats cross-user and over-time interactions as the primary signal for improving skills. SkillClaw continuously aggregates trajectories generated during use and processes them with an autonomous evolver, which identifies recurring behavioral patterns and translates them into updates to the skill set by refining existing skills or extending them with new capabilities. The resulting skills are maintained in a shared repository and synchronized across users, allowing improvements discovered in one context to propagate system-wide while requiring no additional effort from users. By integrating multi-user experience into ongoing skill updates, SkillClaw enables cross-user knowledge transfer and cumulative capability improvement, and experiments on WildClawBench show that limited interaction and feedback, it significantly improves the performance of Qwen3-Max in real-world agent scenarios.</p> - https://arxiv.org/abs/2604.08377 - Thu, 09 Apr 2026 00:00:00 +0000 + Not All Denoising Steps Are Equal: Model Scheduling for Faster Masked Diffusion Language Models + https://arxiv.org/abs/2604.02340 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.02340.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ivan Sedykh, Nikita Sorokin, Valentin Malykh</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Recent advances in masked diffusion language models (MDLMs) narrow the quality gap to autoregressive LMs, but their sampling remains expensive because generation requires many full-sequence denoising passes with a large Transformer and, unlike autoregressive decoding, cannot benefit from KV caching. In this work, we exploit the flexibility of the diffusion framework and study model scheduling, where a smaller MDLM replaces the full model at a subset of denoising steps. Across models trained on OpenWebText and LM1B, we show that early and late denoising steps are substantially more robust to such replacement than middle steps, enabling up to a 17% reduction in FLOPs with only modest degradation in generative perplexity under both unconditional and prefix-conditional generation, while preserving sample diversity. We support these findings with a step-importance analysis based on loss and KL divergence between small and large models across timesteps, as well as an exhaustive search over coarse step segments, both of which identify the middle of the diffusion trajectory as most sensitive consistently across datasets. Our results suggest that simple, architecture-agnostic scheduling rules can significantly accelerate MDLM sampling while largely preserving generation quality.</p> + https://arxiv.org/abs/2604.02340 + Sat, 11 Apr 2026 00:00:00 +0000 + + + Prompt Relay: Inference-Time Temporal Control for Multi-Event Video Generation + https://arxiv.org/abs/2604.10030 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10030.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Gordon Chen, Ziqi Huang, Ziwei Liu</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> Video diffusion models have achieved remarkable progress in generating high-quality videos. However, these models struggle to represent the temporal succession of multiple events in real-world videos and lack explicit mechanisms to control when semantic concepts appear, how long they persist, and the order in which multiple events occur. Such control is especially important for movie-grade video synthesis, where coherent storytelling depends on precise timing, duration, and transitions between events. When using a single paragraph-style prompt to describe a sequence of complex events, models often exhibit semantic entanglement, where concepts intended for different moments in the video bleed into one another, resulting in poor text-video alignment. To address these limitations, we propose Prompt Relay, an inference-time, plug-and-play method to enable fine-grained temporal control in multi-event video generation, requiring no architectural modifications and no additional computational overhead. Prompt Relay introduces a penalty into the cross-attention mechanism, so that each temporal segment attends only to its assigned prompt, allowing the model to represent one semantic concept at a time and thereby improving temporal prompt alignment, reducing semantic interference, and enhancing visual quality.</p> + https://arxiv.org/abs/2604.10030 + Sat, 11 Apr 2026 00:00:00 +0000 + + + Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation + https://arxiv.org/abs/2604.10098 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10098.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zunhai Su, Hengyuan Zhang, Wei Wu, Yifan Zhang, Yaxiu Liu, He Xiao, Qingyao Yang, Yuxuan Sun, Rui Yang, Chao Zhang, Keyu Fan, Weihao Ye, Jing Xiong, Hui Shen, Chaofan Tao, Taiqiang Wu, Zhongwei Wan, Yulei Qian, Yuchen Xie, Ngai Wong</p><p><b>Upvotes:</b> 74</p><p><b>Summary:</b> As the foundational architecture of modern machine learning, Transformers have driven remarkable progress across diverse AI domains. Despite their transformative impact, a persistent challenge across various Transformers is Attention Sink (AS), in which a disproportionate amount of attention is focused on a small subset of specific yet uninformative tokens. AS complicates interpretability, significantly affecting the training and inference dynamics, and exacerbates issues such as hallucinations. In recent years, substantial research has been dedicated to understanding and harnessing AS. However, a comprehensive survey that systematically consolidates AS-related research and offers guidance for future advancements remains lacking. To address this gap, we present the first survey on AS, structured around three key dimensions that define the current research landscape: Fundamental Utilization, Mechanistic Interpretation, and Strategic Mitigation. Our work provides a pivotal contribution by clarifying key concepts and guiding researchers through the evolution and trends of the field. We envision this survey as a definitive resource, empowering researchers and practitioners to effectively manage AS within the current Transformer paradigm, while simultaneously inspiring innovative advancements for the next generation of Transformers. The paper list of this work is available at https://github.com/ZunhaiSu/Awesome-Attention-Sink.</p> + https://arxiv.org/abs/2604.10098 + Sat, 11 Apr 2026 00:00:00 +0000 + + + Rethinking the Diffusion Model from a Langevin Perspective + https://arxiv.org/abs/2604.10465 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10465.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Candi Zheng, Yuan Lan</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> Diffusion models are often introduced from multiple perspectives, such as VAEs, score matching, or flow matching, accompanied by dense and technically demanding mathematics that can be difficult for beginners to grasp. One classic question is: how does the reverse process invert the forward process to generate data from pure noise? This article systematically organizes the diffusion model from a fresh Langevin perspective, offering a simpler, clearer, and more intuitive answer. We also address the following questions: how can ODE-based and SDE-based diffusion models be unified under a single framework? Why are diffusion models theoretically superior to ordinary VAEs? Why is flow matching not fundamentally simpler than denoising or score matching, but equivalent under maximum-likelihood? We demonstrate that the Langevin perspective offers clear and straightforward answers to these questions, bridging existing interpretations of diffusion models, showing how different formulations can be converted into one another within a common framework, and offering pedagogical value for both learners and experienced researchers seeking deeper intuition.</p> + https://arxiv.org/abs/2604.10465 + Sun, 12 Apr 2026 00:00:00 +0000 + + + Tracing the Roots: A Multi-Agent Framework for Uncovering Data Lineage in Post-Training LLMs + https://arxiv.org/abs/2604.10480 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10480.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yu Li, Xiaoran Shang, Qizhi Pei, Yun Zhu, Xin Gao, Honglin Lin, Zhanping Zhong, Zhuoshi Pan, Zheng Liu, Xiaoyang Wang, Conghui He, Dahua Lin, Feng Zhao, Lijun Wu</p><p><b>Upvotes:</b> 20</p><p><b>Summary:</b> Post-training data plays a pivotal role in shaping the capabilities of Large Language Models (LLMs), yet datasets are often treated as isolated artifacts, overlooking the systemic connections that underlie their evolution. To disentangle these complex relationships, we introduce the concept of data lineage to the LLM ecosystem and propose an automated multi-agent framework to reconstruct the evolutionary graph of dataset development. Through large-scale lineage analysis, we characterize domain-specific structural patterns, such as vertical refinement in math-oriented datasets and horizontal aggregation in general-domain corpora. Moreover, we uncover pervasive systemic issues, including structural redundancy induced by implicit dataset intersections and the propagation of benchmark contamination along lineage paths. To demonstrate the practical value of lineage analysis for data construction, we leverage the reconstructed lineage graph to create a lineage-aware diversity-oriented dataset. By anchoring instruction sampling at upstream root sources, this approach mitigates downstream homogenization and hidden redundancy, yielding a more diverse post-training corpus. We further highlight lineage-centric analysis as an efficient and robust topological alternative to sample-level dataset comparison for large-scale data ecosystems. By grounding data construction in explicit lineage structures, our work advances post-training data curation toward a more systematic and controllable paradigm.</p> + https://arxiv.org/abs/2604.10480 + Sun, 12 Apr 2026 00:00:00 +0000 + + + The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents + https://arxiv.org/abs/2604.10577 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10577.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xuwei Ding, Skylar Zhai, Linxin Song, Jiate Li, Taiwei Shi, Nicholas Meade, Siva Reddy, Jian Kang, Jieyu Zhao</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> Computer-use agents (CUAs) can now autonomously complete complex tasks in real digital environments, but when misled, they can also be used to automate harmful actions programmatically. Existing safety evaluations largely target explicit threats such as misuse and prompt injection, but overlook a subtle yet critical setting where user instructions are entirely benign and harm arises from the task context or execution outcome. We introduce OS-BLIND, a benchmark that evaluates CUAs under unintended attack conditions, comprising 300 human-crafted tasks across 12 categories, 8 applications, and 2 threat clusters: environment-embedded threats and agent-initiated harms. Our evaluation on frontier models and agentic frameworks reveals that most CUAs exceed 90% attack success rate (ASR), and even the safety-aligned Claude 4.5 Sonnet reaches 73.0% ASR. More interestingly, this vulnerability becomes even more severe, with ASR rising from 73.0% to 92.7% when Claude 4.5 Sonnet is deployed in multi-agent systems. Our analysis further shows that existing safety defenses provide limited protection when user instructions are benign. Safety alignment primarily activates within the first few steps and rarely re-engages during subsequent execution. In multi-agent systems, decomposed subtasks obscure the harmful intent from the model, causing safety-aligned models to fail. We will release our OS-BLIND to encourage the broader research community to further investigate and address these safety challenges.</p> + https://arxiv.org/abs/2604.10577 + Sun, 12 Apr 2026 00:00:00 +0000 + + + SCOPE: Signal-Calibrated On-Policy Distillation Enhancement with Dual-Path Adaptive Weighting + https://arxiv.org/abs/2604.10688 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10688.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Binbin Zheng, Xing Ma, Yiheng Liang, Jingqing Ruan, Xiaoliang Fu, Kepeng Lin, Benchang Zhu, Ke Zeng, Xunliang Cai</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> On-policy reinforcement learning has become the dominant paradigm for reasoning alignment in large language models, yet its sparse, outcome-level rewards make token-level credit assignment notoriously difficult. On-Policy Distillation (OPD) alleviates this by introducing dense, token-level KL supervision from a teacher model, but typically applies this supervision uniformly across all rollouts, ignoring fundamental differences in signal quality. We propose Signal-Calibrated On-Policy Distillation Enhancement (SCOPE), a dual-path adaptive training framework that routes on-policy rollouts by correctness into two complementary supervision paths. For incorrect trajectories, SCOPE performs teacher-perplexity-weighted KL distillation to prioritize instances where the teacher demonstrates genuine corrective capability, while down-weighting unreliable guidance. For correct trajectories, it applies student-perplexity-weighted MLE to concentrate reinforcement on low-confidence samples at the capability boundary rather than over-reinforcing already mastered ones. Both paths employ a group-level normalization to adaptively calibrate weight distributions, accounting for the intrinsic difficulty variance across prompts. Extensive experiments on six reasoning benchmarks show that SCOPE achieves an average relative improvement of 11.42% in Avg@32 and 7.30% in Pass@32 over competitive baselines, demonstrating its consistent effectiveness.</p> + https://arxiv.org/abs/2604.10688 + Sun, 12 Apr 2026 00:00:00 +0000 + + + Audio-Omni: Extending Multi-modal Understanding to Versatile Audio Generation and Editing + https://arxiv.org/abs/2604.10708 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10708.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zeyue Tian, Binxin Yang, Zhaoyang Liu, Jiexuan Zhang, Ruibin Yuan, Hubery Yin, Qifeng Chen, Chen Li, Jing Lv, Wei Xue, Yike Guo</p><p><b>Upvotes:</b> 40</p><p><b>Summary:</b> Recent progress in multimodal models has spurred rapid advances in audio understanding, generation, and editing. However, these capabilities are typically addressed by specialized models, leaving the development of a truly unified framework that can seamlessly integrate all three tasks underexplored. While some pioneering works have explored unifying audio understanding and generation, they often remain confined to specific domains. To address this, we introduce Audio-Omni, the first end-to-end framework to unify generation and editing across general sound, music, and speech domains, with integrated multi-modal understanding capabilities. Our architecture synergizes a frozen Multimodal Large Language Model for high-level reasoning with a trainable Diffusion Transformer for high-fidelity synthesis. To overcome the critical data scarcity in audio editing, we construct AudioEdit, a new large-scale dataset comprising over one million meticulously curated editing pairs. Extensive experiments demonstrate that Audio-Omni achieves state-of-the-art performance across a suite of benchmarks, outperforming prior unified approaches while achieving performance on par with or superior to specialized expert models. Beyond its core capabilities, Audio-Omni exhibits remarkable inherited capabilities, including knowledge-augmented reasoning generation, in-context generation, and zero-shot cross-lingual control for audio generation, highlighting a promising direction toward universal generative audio intelligence. The code, model, and dataset will be publicly released on https://zeyuet.github.io/Audio-Omni.</p> + https://arxiv.org/abs/2604.10708 + Sun, 12 Apr 2026 00:00:00 +0000 + + + Narrative-Driven Paper-to-Slide Generation via ArcDeck + https://arxiv.org/abs/2604.11969 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11969.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tarik Can Ozden, Sachidanand VS, Furkan Horoz, Ozgur Kara, Junho Kim, James Matthew Rehg</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> We introduce ArcDeck, a multi-agent framework that formulates paper-to-slide generation as a structured narrative reconstruction task. Unlike existing methods that directly summarize raw text into slides, ArcDeck explicitly models the source paper's logical flow. It first parses the input to construct a discourse tree and establish a global commitment document, ensuring the high-level intent is preserved. These structural priors then guide an iterative multi-agent refinement process, where specialized agents iteratively critique and revise the presentation outline before rendering the final visual layouts and designs. To evaluate our approach, we also introduce ArcBench, a newly curated benchmark of academic paper-slide pairs. Experimental results demonstrate that explicit discourse modeling, combined with role-specific agent coordination, significantly improves the narrative flow and logical coherence of the generated presentations.</p> + https://arxiv.org/abs/2604.11969 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Do Thought Streams Matter? Evaluating Reasoning in Gemini Vision-Language Models for Video Scene Understanding + https://arxiv.org/abs/2604.11177 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11177.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shivam Sharma, Sankalp Nagaonkar, Ashish Choithani, Ashutosh Trivedi</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> We benchmark how internal reasoning traces, which we call thought streams, affect video scene understanding in vision-language models. Using four configurations of Google's Gemini 2.5 Flash and Flash Lite across scenes extracted from 100 hours of video, we ask three questions: does more thinking lead to better outputs, where do the gains stop, and what do these models actually think about? We introduce three evaluation metrics. Contentfulness measures how much of the thought stream is useful scene content versus meta-commentary. Thought-Final Coverage measures how faithfully the thought stream translates into the final output. Dominant Entity Analysis identifies which subjects, actions, and settings the model focuses on. GPT-5 serves as an independent judge. We find that quality gains from additional thinking plateau quickly, with most improvement occurring in the first few hundred tokens. Flash Lite offers the best balance between quality and token usage. Tight reasoning budgets cause the model to add content in the final output that it never reasoned about, a form of compression-step hallucination. Despite being different model tiers, Flash and Flash Lite produce similar thought streams, though they differ in style: Flash discusses its reasoning process, while Lite focuses on describing the scene.</p> + https://arxiv.org/abs/2604.11177 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Representations Before Pixels: Semantics-Guided Hierarchical Video Prediction + https://arxiv.org/abs/2604.11707 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11707.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Efstathios Karypidis, Spyros Gidaris, Nikos Komodakis</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Accurate future video prediction requires both high visual fidelity and consistent scene semantics, particularly in complex dynamic environments such as autonomous driving. We present Re2Pix, a hierarchical video prediction framework that decomposes forecasting into two stages: semantic representation prediction and representation-guided visual synthesis. Instead of directly predicting future RGB frames, our approach first forecasts future scene structure in the feature space of a frozen vision foundation model, and then conditions a latent diffusion model on these predicted representations to render photorealistic frames. This decomposition enables the model to focus first on scene dynamics and then on appearance generation. A key challenge arises from the train-test mismatch between ground-truth representations available during training and predicted ones used at inference. To address this, we introduce two conditioning strategies, nested dropout and mixed supervision, that improve robustness to imperfect autoregressive predictions. Experiments on challenging driving benchmarks demonstrate that the proposed semantics-first design significantly improves temporal semantic consistency, perceptual quality, and training efficiency compared to strong diffusion baselines. We provide the implementation code at https://github.com/Sta8is/Re2Pix</p> + https://arxiv.org/abs/2604.11707 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Continuous Adversarial Flow Models + https://arxiv.org/abs/2604.11521 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11521.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shanchuan Lin, Ceyuan Yang, Zhijie Lin, Hao Chen, Haoqi Fan</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> We propose continuous adversarial flow models, a type of continuous-time flow model trained with an adversarial objective. Unlike flow matching, which uses a fixed mean-squared-error criterion, our approach introduces a learned discriminator to guide training. This change in objective induces a different generalized distribution, which empirically produces samples that are better aligned with the target data distribution. Our method is primarily proposed for post-training existing flow-matching models, although it can also train models from scratch. On the ImageNet 256px generation task, our post-training substantially improves the guidance-free FID of latent-space SiT from 8.26 to 3.63 and of pixel-space JiT from 7.17 to 3.57. It also improves guided generation, reducing FID from 2.06 to 1.53 for SiT and from 1.86 to 1.80 for JiT. We further evaluate our approach on text-to-image generation, where it achieves improved results on both the GenEval and DPG benchmarks.</p> + https://arxiv.org/abs/2604.11521 + Mon, 13 Apr 2026 00:00:00 +0000 + + + General365: Benchmarking General Reasoning in Large Language Models Across Diverse and Challenging Tasks + https://arxiv.org/abs/2604.11778 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11778.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Junlin Liu, Shengnan An, Shuang Zhou, Dan Ma, Shixiong Luo, Ying Xie, Yuan Zhang, Wenling Yuan, Yifan Zhou, Xiaoyu Li, Ziwen Wang, Xuezhi Cao, Xunliang Cai</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Contemporary large language models (LLMs) have demonstrated remarkable reasoning capabilities, particularly in specialized domains like mathematics and physics. However, their ability to generalize these reasoning skills to more general and broader contexts--often termed general reasoning--remains under-explored. Unlike domain-specific reasoning, general reasoning relies less on expert knowledge but still presents formidable reasoning challenges, such as complex constraints, nested logical branches, and semantic interference. To address this gap, we introduce General365, a benchmark specifically designed to assess general reasoning in LLMs. By restricting background knowledge to a K-12 level, General365 explicitly decouples reasoning from specialized expertise. The benchmark comprises 365 seed problems and 1,095 variant problems across eight categories, ensuring both high difficulty and diversity. Evaluations across 26 leading LLMs reveal that even the top-performing model achieves only 62.8% accuracy, in stark contrast to the near-perfect performances of LLMs in math and physics benchmarks. These results suggest that the reasoning abilities of current LLMs are heavily domain-dependent, leaving significant room for improvement in broader applications. We envision General365 as a catalyst for advancing LLM reasoning beyond domain-specific tasks toward robust, general-purpose real-world scenarios. Code, Dataset, and Leaderboard: https://general365.github.io</p> + https://arxiv.org/abs/2604.11778 + Mon, 13 Apr 2026 00:00:00 +0000 + + + You Only Judge Once: Multi-response Reward Modeling in a Single Forward Pass + https://arxiv.org/abs/2604.10966 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10966.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yinuo Yang, Zixian Ma, Manasi Ganti, Jieyu Zhang, Ranjay Krishna</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> We present a discriminative multimodal reward model that scores all candidate responses in a single forward pass. Conventional discriminative reward models evaluate each response independently, requiring multiple forward passes, one for each potential response. Our approach concatenates multiple responses with separator tokens and applies cross-entropy over their scalar scores, enabling direct comparative reasoning and efficient N-way preference learning. The multi-response design also yields up to Ntimes wall-clock speedup and FLOPs reduction over conventional single-response scoring. To enable N-way reward evaluation beyond existing pairwise benchmarks, we construct two new benchmarks: (1) MR^2Bench-Image contains human-annotated rankings over responses from 8 diverse models; (2) MR^2Bench-Video is a large-scale video-based reward benchmark derived from 94K crowdsourced pairwise human judgments over video question-answering spanning 19 models, denoised via preference graph ensemble. Both benchmarks provide 4-response evaluation variants sampled from the full rankings. Built on a 4B vision-language backbone with LoRA fine-tuning and a lightweight MLP value head, our model achieves state-of-the-art results on six multimodal reward benchmarks, including MR^2Bench-Image, MR^2Bench-Video, and four other existing benchmarks. Our model outperforms existing larger generative and discriminative reward models. We further demonstrate that our reward model, when used in reinforcement learning with GRPO, produces improved policy models that maintain performance across standard multimodal benchmarks while substantially improving open-ended generation quality, outperforming a single-response discriminative reward model (RM) baseline by a large margin in both training stability and open-ended generation quality.</p> + https://arxiv.org/abs/2604.10966 + Mon, 13 Apr 2026 00:00:00 +0000 + + + LARY: A Latent Action Representation Yielding Benchmark for Generalizable Vision-to-Action Alignment + https://arxiv.org/abs/2604.11689 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11689.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dujun Nie, Fengjiao Chen, Qi Lv, Jun Kuang, Xiaoyu Li, Xuezhi Cao, Xunliang Cai</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> While the shortage of explicit action data limits Vision-Language-Action (VLA) models, human action videos offer a scalable yet unlabeled data source. A critical challenge in utilizing large-scale human video datasets lies in transforming visual signals into ontology-independent representations, known as latent actions. However, the capacity of latent action representation to derive robust control from visual observations has yet to be rigorously evaluated. We introduce the Latent Action Representation Yielding (LARY) Benchmark, a unified framework for evaluating latent action representations on both high-level semantic actions (what to do) and low-level robotic control (how to do). The comprehensively curated dataset encompasses over one million videos (1,000 hours) spanning 151 action categories, alongside 620K image pairs and 595K motion trajectories across diverse embodiments and environments. Our experiments reveal two crucial insights: (i) General visual foundation models, trained without any action supervision, consistently outperform specialized embodied latent action models. (ii) Latent-based visual space is fundamentally better aligned to physical action space than pixel-based space. These results suggest that general visual representations inherently encode action-relevant knowledge for physical control, and that semantic-level abstraction serves as a fundamentally more effective pathway from vision to action than pixel-level reconstruction.</p> + https://arxiv.org/abs/2604.11689 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Anthropogenic Regional Adaptation in Multimodal Vision-Language Model + https://arxiv.org/abs/2604.11490 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11490.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Samuel Cahyawijaya, Peerat Limkonchotiwat, Tack Hwa Wong, Hitesh Laxmichand Patel, Amit Agarwal, Manuel Antonio Rufino, Carlos Rafael Catalan, Muhammad Reza Qorib, Vicky Feliren, Holy Lovenia, Aye Hninn Khine, Frederikus Hudi, David Anugraha, Alham Fikri Aji, Romrawin Chumpu, Viet-Thanh Pham, Minghan Wang, Mohamed Fazli Imam, Ruochen Zhang, Joseph Marvin Imperial, Do Xuan Long, Musa Izzanardi Wijanarko, Joel Ruben Antony Moniz, Patrick Amadeus Irawan, Hanif Muhammad Zhafran, Isaiah Flores, Ira Salsabila, Jun Kevin, Jostin Jerico Rosal, Patricia Nicole Monderin, Kun Kerdthaisong, Ahmad Mustafid, My Chiffon Nguyen, Natchapon Jongwiriyanurak, Siva Worajitwannakul, Haochen Li, Adrian Xuan Wei Lim, Bin Wang, Muhammad Ravi Shulthan Habibi, Lynnette Hui Xian Ng, Mithil Bangera, Yeshil Bangera, Priyaranjan Pattnayak, Dun Li Chan, Sherissa Caren Djuniwar, Hee Ming Shan</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> While the field of vision-language (VL) has achieved remarkable success in integrating visual and textual information across multiple languages and domains, there is still no dedicated framework for assessing human-centric alignment in vision-language systems. We offer two contributions to address this gap. First, we introduce Anthropogenic Regional Adaptation: a novel paradigm that aims to optimize model relevance to specific regional contexts while ensuring the retention of global generalization capabilities. Second, we present a simple, but effective adaptation method named Geographical-generalization-made-easy (GG-EZ), which utilizes regional data filtering and model merging. Through comprehensive experiments on 3 VL architectures: large vision-language models, text-to-image diffusion models, and vision-language embedding models, and a case study in Southeast Asia (SEA) regional adaptation, we demonstrate the importance of Anthropogenic Regional Adaptation and the effectiveness of GG-EZ, showing 5-15% gains in cultural relevance metrics across SEA while maintaining over 98% of global performance and even occasionally surpassing it. Our findings establish Anthropogenic Regional Alignment as a foundational paradigm towards applicability of multimodal vision-language models in diverse regions and demonstrate a simple-yet-effective baseline method that optimizes regional value alignment while preserving global generalization.</p> + https://arxiv.org/abs/2604.11490 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Mobile GUI Agent Privacy Personalization with Trajectory Induced Preference Optimization + https://arxiv.org/abs/2604.11259 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11259.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhixin Lin, Jungang Li, Dongliang Xu, Shidong Pan, Yibo Shi, Yuchi Liu, Yuecong Min, Yue Yao</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Mobile GUI agents powered by Multimodal Large Language Models (MLLMs) can execute complex tasks on mobile devices. Despite this progress, most existing systems still optimize task success or efficiency, neglecting users' privacy personalization. In this paper, we study the often-overlooked problem of agent personalization. We observe that personalization can induce systematic structural heterogeneity in execution trajectories. For example, privacy-first users often prefer protective actions, e.g., refusing permissions, logging out, and minimizing exposure, leading to logically different execution trajectories from utility-first users. Such variable-length and structurally different trajectories make standard preference optimization unstable and less informative. To address this issue, we propose Trajectory Induced Preference Optimization (TIPO), which uses preference-intensity weighting to emphasize key privacy-related steps and padding gating to suppress alignment noise. Results on our Privacy Preference Dataset show that TIPO improves persona alignment and distinction while preserving strong task executability, achieving 65.60% SR, 46.22 Compliance, and 66.67% PD, outperforming existing optimization methods across various GUI tasks. The code and dataset will be publicly released at https://github.com/Zhixin-L/TIPO.</p> + https://arxiv.org/abs/2604.11259 + Mon, 13 Apr 2026 00:00:00 +0000 + + + From Reasoning to Agentic: Credit Assignment in Reinforcement Learning for Large Language Models + https://arxiv.org/abs/2604.09459 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09459.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenchen Zhang</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Reinforcement learning (RL) for large language models (LLMs) increasingly relies on sparse, outcome-level rewards -- yet determining which actions within a long trajectory caused the outcome remains difficult. This credit assignment (CA) problem manifests in two regimes: reasoning RL, where credit must be distributed across tokens and steps within a single chain-of-thought generation (500--30K+ tokens); and agentic RL, where multi-turn environment interaction introduces stochastic transitions, partial observability, and horizons of 100+ turns (100K--1M tokens), making episode-level credit increasingly uninformative. We survey 47 CA methods (41 core, 6 adjacent enablers) published between 2024 and early 2026, organizing them in a two-dimensional taxonomy by assignment granularity (token, segment, step, turn, multi-agent) and methodology (Monte Carlo, temporal difference, model-based, game-theoretic, information-theoretic). Beyond the survey itself, we contribute three reusable resources: (1) a structured, machine-readable paper inventory with taxonomy labels, baseline families, and evidence levels; (2) a reporting checklist for future CA papers, validated against the reviewed literature to identify systematic methodological gaps; and (3) a benchmark protocol specification with task families, metadata requirements, and controlled bifurcation tasks, accompanied by a method selection decision tree. Our synthesis suggests that the shift from reasoning to agentic RL complicates and reshapes the credit assignment landscape: reasoning CA is maturing around process reward models and critic-free group comparison, while agentic CA is driving genuinely new approaches -- hindsight counterfactual analysis, privileged asymmetric critics, and turn-level MDP reformulations -- that have no direct precedent in reasoning RL.</p> + https://arxiv.org/abs/2604.09459 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Tasks + https://arxiv.org/abs/2604.11753 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11753.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yoonsang Lee, Howard Yen, Xi Ye, Danqi Chen</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> We study parallel test-time scaling for long-horizon agentic tasks such as agentic search and deep research, where multiple rollouts are generated in parallel and aggregated into a final response. While such scaling has proven effective for chain-of-thought reasoning, agentic tasks pose unique challenges: trajectories are long, multi-turn, and tool-augmented, and outputs are often open-ended. Aggregating only final answers discards rich information from trajectories, while concatenating all trajectories exceeds the model's context window. To address this, we propose AggAgent, an aggregation agent that treats parallel trajectories as an environment. We equip it with lightweight tools to inspect candidate solutions and search across trajectories, enabling it to navigate and synthesize information on demand. Across six benchmarks and three model families (GLM-4.7, Qwen3.5, MiniMax-M2.5), AggAgent outperforms all existing aggregation methods-by up to 5.3% absolute on average and 10.3% on two deep research tasks-while adding minimal overhead, as the aggregation cost remains bounded by a single agentic rollout. Our findings establish agentic aggregation as an effective and cost-efficient approach to parallel test-time scaling.</p> + https://arxiv.org/abs/2604.11753 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Solving Physics Olympiad via Reinforcement Learning on Physics Simulators + https://arxiv.org/abs/2604.11805 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11805.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mihir Prabhudesai, Aryan Satpathy, Yangmin Li, Zheyang Qin, Nikash Bhardwaj, Amir Zadeh, Chuan Li, Katerina Fragkiadaki, Deepak Pathak</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> We have witnessed remarkable advances in LLM reasoning capabilities with the advent of DeepSeek-R1. However, much of this progress has been fueled by the abundance of internet question-answer (QA) pairs, a major bottleneck going forward, since such data is limited in scale and concentrated mainly in domains like mathematics. In contrast, other sciences such as physics lack large-scale QA datasets to effectively train reasoning-capable models. In this work, we show that physics simulators can serve as a powerful alternative source of supervision for training LLMs for physical reasoning. We generate random scenes in physics engines, create synthetic question-answer pairs from simulated interactions, and train LLMs using reinforcement learning on this synthetic data. Our models exhibit zero-shot sim-to-real transfer to real-world physics benchmarks: for example, training solely on synthetic simulated data improves performance on IPhO (International Physics Olympiad) problems by 5-10 percentage points across model sizes. These results demonstrate that physics simulators can act as scalable data generators, enabling LLMs to acquire deep physical reasoning skills beyond the limitations of internet-scale QA data. Code available at: https://sim2reason.github.io/.</p> + https://arxiv.org/abs/2604.11805 + Mon, 13 Apr 2026 00:00:00 +0000 + + + SemaClaw: A Step Towards General-Purpose Personal AI Agents through Harness Engineering + https://arxiv.org/abs/2604.11548 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11548.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ningyan Zhu, Huacan Wang, Jie Zhou, Feiyu Chen, Shuo Zhang, Ge Chen, Chen Liu, Jiarou Wu, Wangyi Chen, Xiaofeng Mou, Yi Xu</p><p><b>Upvotes:</b> 18</p><p><b>Summary:</b> The rise of OpenClaw in early 2026 marks the moment when millions of users began deploying personal AI agents into their daily lives, delegating tasks ranging from travel planning to multi-step research. This scale of adoption signals that two parallel arcs of development have reached an inflection point. First is a paradigm shift in AI engineering, evolving from prompt and context engineering to harness engineering-designing the complete infrastructure necessary to transform unconstrained agents into controllable, auditable, and production-reliable systems. As model capabilities converge, this harness layer is becoming the primary site of architectural differentiation. Second is the evolution of human-agent interaction from discrete tasks toward a persistent, contextually aware collaborative relationship, which demands open, trustworthy and extensible harness infrastructure. We present SemaClaw, an open-source multi-agent application framework that addresses these shifts by taking a step towards general-purpose personal AI agents through harness engineering. Our primary contributions include a DAG-based two-phase hybrid agent team orchestration method, a PermissionBridge behavioral safety system, a three-tier context management architecture, and an agentic wiki skill for automated personal knowledge base construction.</p> + https://arxiv.org/abs/2604.11548 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Introspective Diffusion Language Models + https://arxiv.org/abs/2604.11035 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11035.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yifan Yu, Yuqing Jian, Junxiong Wang, Zhongzhu Zhou, Donglin Zhuang, Xinyu Fang, Sri Yanamandra, Xiaoxia Wu, Qingyang Wu, Shuaiwen Leon Song, Tri Dao, Ben Athiwaratkun, James Zou, Fan Lai, Chenfeng Xu</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> Diffusion language models promise parallel generation, yet still lag behind autoregressive (AR) models in quality. We stem this gap to a failure of introspective consistency: AR models agree with their own generations, while DLMs often do not. We define the introspective acceptance rate, which measures whether a model accepts its previously generated tokens. This reveals why AR training has a structural advantage: causal masking and logit shifting implicitly enforce introspective consistency. Motivated by this observation, we introduce Introspective Diffusion Language Model (I-DLM), a paradigm that retains diffusion-style parallel decoding while inheriting the introspective consistency of AR training. I-DLM uses a novel introspective strided decoding (ISD) algorithm, which enables the model to verify previously generated tokens while advancing new ones in the same forward pass. From a systems standpoint, we build I-DLM inference engine on AR-inherited optimizations and further customize it with a stationary-batch scheduler. To the best of our knowledge, I-DLM is the first DLM to match the quality of its same-scale AR counterpart while outperforming prior DLMs in both model quality and practical serving efficiency across 15 benchmarks. It reaches 69.6 on AIME-24 and 45.7 on LiveCodeBench-v6, exceeding LLaDA-2.1-mini (16B) by more than 26 and 15 points, respectively. Beyond quality, I-DLM is designed for the growing demand of large-concurrency serving, delivering about 3x higher throughput than prior state-of-the-art DLMs.</p> + https://arxiv.org/abs/2604.11035 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Sema Code: Decoupling AI Coding Agents into Programmable, Embeddable Infrastructure + https://arxiv.org/abs/2604.11045 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11045.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Huacan Wang, Jie Zhou, Ningyan Zhu, Shuo Zhang, Feiyu Chen, Jiarou Wu, Ge Chen, Chen Liu, Wangyi Chen, Xiaofeng Mou, Yi Xu</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> AI coding agents have become central to developer workflows, yet every existing solution locks its reasoning capabilities within a specific delivery form, such as a CLI, IDE plugin, or web application. This limitation creates systemic barriers when enterprises attempt to reuse these capabilities across heterogeneous engineering environments. To address this challenge, we present Sema Code, an open AI coding framework built on the principle of being embeddable, pluggable, and framework-first. Sema Code completely decouples the core agent engine from all client layers, publishing it as a standalone npm library that any runtime can drive programmatically. Built around this architecture, we designed eight key mechanisms: multi-tenant engine isolation, FIFO input queuing with safe session reconstruction, adaptive context compression, multi-agent collaborative scheduling, intelligent Todo-based process management, four-layer asynchronous permission control, three-tier ecosystem integration spanning MCP, Skills, and Plugins, and a background task framework with separated execution and observation privileges. These mechanisms collectively address the engineering challenges of transforming a complex agent engine into a shared, programmable core. Demonstrating its architectural versatility, the same Sema Core engine simultaneously powers a VSCode extension and a multi-channel messaging gateway, which we name SemaClaw, to unify agent interactions across platforms such as Telegram and Feishu. These represent two fundamentally different product forms sharing an identical reasoning kernel, differing only at the client layer.</p> + https://arxiv.org/abs/2604.11045 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Audio Flamingo Next: Next-Generation Open Audio-Language Models for Speech, Sound, and Music + https://arxiv.org/abs/2604.10905 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10905.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sreyan Ghosh, Arushi Goel, Kaousheik Jayakumar, Lasha Koroshinadze, Nishit Anand, Zhifeng Kong, Siddharth Gururani, Sang-gil Lee, Jaehyeon Kim, Aya Aljafari, Chao-Han Huck Yang, Sungwon Kim, Ramani Duraiswami, Dinesh Manocha, Mohammad Shoeybi, Bryan Catanzaro, Ming-Yu Liu, Wei Ping</p><p><b>Upvotes:</b> 28</p><p><b>Summary:</b> We present Audio Flamingo Next (AF-Next), the next-generation and most capable large audio-language model in the Audio Flamingo series, designed to advance understanding and reasoning over speech, environmental sounds and music. Compared to Audio Flamingo 3, AF-Next introduces: (i) a stronger foundational audio-language model that significantly improves accuracy across diverse audio understanding tasks; (ii) scalable strategies for constructing large-scale audio understanding and reasoning data beyond existing academic benchmarks; (iii) support for long and complex audio inputs up to 30 minutes; and (iv) Temporal Audio Chain-of-Thought, a new reasoning paradigm that explicitly grounds intermediate reasoning steps to timestamps in long audio, enabling fine-grained temporal alignment and improved interpretability. To enable these capabilities, we first conduct a systematic analysis of Audio Flamingo 3 to identify key gaps in audio understanding and reasoning. We then curate and scale new large-scale datasets totaling over 1 million hours to address these limitations and expand the existing AudioSkills-XL, LongAudio-XL, AF-Think and AF-Chat datasets. AF-Next is trained using a curriculum-based strategy spanning pre-training, mid-training and post-training stages. Extensive experiments across 20 audio understanding and reasoning benchmarks, including challenging long-audio tasks, show that AF-Next outperforms similarly sized open models by large margins and remains highly competitive with and sometimes surpasses, much larger open-weight and closed models. Beyond benchmark performance, AF-Next exhibits strong real-world utility and transfers well to unseen tasks, highlighting its robustness and generalization ability. In addition to all data, code and methods, we open-source 3 variants of AF-Next, including AF-Next-Instruct, AF-Next-Think and AF-Next-Captioner.</p> + https://arxiv.org/abs/2604.10905 + Mon, 13 Apr 2026 00:00:00 +0000 + + + CocoaBench: Evaluating Unified Digital Agents in the Wild + https://arxiv.org/abs/2604.11201 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11201.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> CocoaBench Team, Shibo Hao, Zhining Zhang, Zhiqi Liang, Tianyang Liu, Yuheng Zha, Qiyue Gao, Jixuan Chen, Zilong Wang, Zhoujun Cheng, Haoxiang Zhang, Junli Wang, Hexi Jin, Boyuan Zheng, Kun Zhou, Yu Wang, Feng Yao, Licheng Liu, Yijiang Li, Zhifei Li, Zhengtao Han, Pracha Promthaw, Tommaso Cerruti, Xiaohan Fu, Ziqiao Ma, Jingbo Shang, Lianhui Qin, Julian McAuley, Eric P. Xing, Zhengzhong Liu, Rupesh Kumar Srivastava, Zhiting Hu</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> LLM agents now perform strongly in software engineering, deep research, GUI automation, and various other applications, while recent agent scaffolds and models are increasingly integrating these capabilities into unified systems. Yet, most evaluations still test these capabilities in isolation, which leaves a gap for more diverse use cases that require agents to combine different capabilities. We introduce CocoaBench, a benchmark for unified digital agents built from human-designed, long-horizon tasks that require flexible composition of vision, search, and coding. Tasks are specified only by an instruction and an automatic evaluation function over the final output, enabling reliable and scalable evaluation across diverse agent infrastructures. We also present CocoaAgent, a lightweight shared scaffold for controlled comparison across model backbones. Experiments show that current agents remain far from reliable on CocoaBench, with the best evaluated system achieving only 45.1% success rate. Our analysis further points to substantial room for improvement in reasoning and planning, tool use and execution, and visual grounding.</p> + https://arxiv.org/abs/2604.11201 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Pseudo-Unification: Entropy Probing Reveals Divergent Information Patterns in Unified Multimodal Models + https://arxiv.org/abs/2604.10949 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10949.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Songlin Yang, Xianghao Kong, Anyi Rao</p><p><b>Upvotes:</b> 39</p><p><b>Summary:</b> Unified multimodal models (UMMs) were designed to combine the reasoning ability of large language models (LLMs) with the generation capability of vision models. In practice, however, this synergy remains elusive: UMMs fail to transfer LLM-like reasoning to image synthesis and exhibit divergent response behaviors. We term this phenomenon pseudo-unification. Diagnosing its internal causes is important, but existing probing methods either lack model-internal insight or ignore prompt-response dependencies. To address these limitations, we propose an information-theoretic probing framework that jointly analyzes how UMMs encode inputs and generate outputs. Applied to ten representative UMMs, our framework reveals that pseudo-unification stems from a dual divergence: (i) Modality-Asymmetric Encoding, where vision and language follow different entropy trajectories, and (ii) Pattern-Split Response, where text generation exhibits high-entropy creativity while image synthesis enforces low-entropy fidelity. Only models that unify both sides (e.g., via contextual prediction) achieve more genuine unification, enabling stronger reasoning-based text-to-image generation even with fewer parameters. Our work provides the first model-internal probing of unification, demonstrating that real multimodal synergy requires consistency in information flow, not just shared parameters.</p> + https://arxiv.org/abs/2604.10949 + Mon, 13 Apr 2026 00:00:00 +0000 + + + CodeTracer: Towards Traceable Agent States + https://arxiv.org/abs/2604.11641 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11641.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Han Li, Yifan Yao, Letian Zhu, Rili Feng, Hongyi Ye, Jiaming Wang, Yancheng He, Pengyu Zou, Lehan Zhang, Xinping Lei, Haoyang Huang, Ken Deng, Ming Sun, Zhaoxiang Zhang, He Ye, Jiaheng Liu</p><p><b>Upvotes:</b> 39</p><p><b>Summary:</b> Code agents are advancing rapidly, but debugging them is becoming increasingly difficult. As frameworks orchestrate parallel tool calls and multi-stage workflows over complex tasks, making the agent's state transitions and error propagation hard to observe. In these runs, an early misstep can trap the agent in unproductive loops or even cascade into fundamental errors, forming hidden error chains that make it hard to tell when the agent goes off track and why. Existing agent tracing analyses either focus on simple interaction or rely on small-scale manual inspection, which limits their scalability and usefulness for real coding workflows. We present CodeTracer, a tracing architecture that parses heterogeneous run artifacts through evolving extractors, reconstructs the full state transition history as a hierarchical trace tree with persistent memory, and performs failure onset localization to pinpoint the failure origin and its downstream chain. To enable systematic evaluation, we construct CodeTraceBench from a large collection of executed trajectories generated by four widely used code agent frameworks on diverse code tasks (e.g., bug fixing, refactoring, and terminal interaction), with supervision at both the stage and step levels for failure localization. Experiments show that CodeTracer substantially outperforms direct prompting and lightweight baselines, and that replaying its diagnostic signals consistently recovers originally failed runs under matched budgets. Our code and data are publicly available.</p> + https://arxiv.org/abs/2604.11641 + Mon, 13 Apr 2026 00:00:00 +0000 + + + OccuBench: Evaluating AI Agents on Real-World Professional Tasks via Language World Models + https://arxiv.org/abs/2604.10866 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.10866.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiaomeng Hu, Yinger Zhang, Fei Huang, Jianhong Tu, Yang Su, Lianghao Deng, Yuxuan Liu, Yantao Liu, Dayiheng Liu, Tsung-Yi Ho</p><p><b>Upvotes:</b> 60</p><p><b>Summary:</b> AI agents are expected to perform professional work across hundreds of occupational domains (from emergency department triage to nuclear reactor safety monitoring to customs import processing), yet existing benchmarks can only evaluate agents in the few domains where public environments exist. We introduce OccuBench, a benchmark covering 100 real-world professional task scenarios across 10 industry categories and 65 specialized domains, enabled by Language World Models (LWMs) that simulate domain-specific environments through LLM-driven tool response generation. Our multi-agent synthesis pipeline automatically produces evaluation instances with guaranteed solvability, calibrated difficulty, and document-grounded diversity. OccuBench evaluates agents along two complementary dimensions: task completion across professional domains and environmental robustness under controlled fault injection (explicit errors, implicit data degradation, and mixed faults). We evaluate 15 frontier models across 8 model families and find that: (1) no single model dominates all industries, as each has a distinct occupational capability profile; (2) implicit faults (truncated data, missing fields) are harder than both explicit errors (timeouts, 500s) and mixed faults, because they lack overt error signals and require the agent to independently detect data degradation; (3) larger models, newer generations, and higher reasoning effort consistently improve performance. GPT-5.2 improves by 27.5 points from minimal to maximum reasoning effort; and (4) strong agents are not necessarily strong environment simulators. Simulator quality is critical for LWM-based evaluation reliability. OccuBench provides the first systematic cross-industry evaluation of AI agents on professional occupational tasks.</p> + https://arxiv.org/abs/2604.10866 + Mon, 13 Apr 2026 00:00:00 +0000 + + + OmniShow: Unifying Multimodal Conditions for Human-Object Interaction Video Generation + https://arxiv.org/abs/2604.11804 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11804.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Donghao Zhou, Guisheng Liu, Hao Yang, Jiatong Li, Jingyu Lin, Xiaohu Huang, Yichen Liu, Xin Gao, Cunjian Chen, Shilei Wen, Chi-Wing Fu, Pheng-Ann Heng</p><p><b>Upvotes:</b> 69</p><p><b>Summary:</b> In this work, we study Human-Object Interaction Video Generation (HOIVG), which aims to synthesize high-quality human-object interaction videos conditioned on text, reference images, audio, and pose. This task holds significant practical value for automating content creation in real-world applications, such as e-commerce demonstrations, short video production, and interactive entertainment. However, existing approaches fail to accommodate all these requisite conditions. We present OmniShow, an end-to-end framework tailored for this practical yet challenging task, capable of harmonizing multimodal conditions and delivering industry-grade performance. To overcome the trade-off between controllability and quality, we introduce Unified Channel-wise Conditioning for efficient image and pose injection, and Gated Local-Context Attention to ensure precise audio-visual synchronization. To effectively address data scarcity, we develop a Decoupled-Then-Joint Training strategy that leverages a multi-stage training process with model merging to efficiently harness heterogeneous sub-task datasets. Furthermore, to fill the evaluation gap in this field, we establish HOIVG-Bench, a dedicated and comprehensive benchmark for HOIVG. Extensive experiments demonstrate that OmniShow achieves overall state-of-the-art performance across various multimodal conditioning settings, setting a solid standard for the emerging HOIVG task.</p> + https://arxiv.org/abs/2604.11804 + Mon, 13 Apr 2026 00:00:00 +0000 + + + RationalRewards: Reasoning Rewards Scale Visual Generation Both Training and Test Time + https://arxiv.org/abs/2604.11626 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11626.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haozhe Wang, Cong Wei, Weiming Ren, Jiaming Liu, Fangzhen Lin, Wenhu Chen</p><p><b>Upvotes:</b> 100</p><p><b>Summary:</b> Most reward models for visual generation reduce rich human judgments to a single unexplained score, discarding the reasoning that underlies preference. We show that teaching reward models to produce explicit, multi-dimensional critiques before scoring transforms them from passive evaluators into active optimization tools, improving generators in two complementary ways: at training time, structured rationales provide interpretable, fine-grained rewards for reinforcement learning; at test time, a Generate-Critique-Refine loop turns critiques into targeted prompt revisions that improve outputs without any parameter updates. To train such a reward model without costly rationale annotations, we introduce Preference-Anchored Rationalization (PARROT), a principled framework that recovers high-quality rationales from readily available preference data through anchored generation, consistency filtering, and distillation. The resulting model, RationalRewards (8B), achieves state-of-the-art preference prediction among open-source reward models, competitive with Gemini-2.5-Pro, while using 10-20x less training data than comparable baselines. As an RL reward, it consistently improves text-to-image and image-editing generators beyond scalar alternatives. Most strikingly, its test-time critique-and-refine loop matches or exceeds RL-based fine-tuning on several benchmarks, suggesting that structured reasoning can unlock latent capabilities in existing generators that suboptimal prompts fail to elicit.</p> + https://arxiv.org/abs/2604.11626 + Mon, 13 Apr 2026 00:00:00 +0000 + + + The Past Is Not Past: Memory-Enhanced Dynamic Reward Shaping + https://arxiv.org/abs/2604.11297 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11297.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yang Liu, Enxi Wang, Yufei Gao, Weixin Zhang, Bo Wang, Zhiyuan Zeng, Yikai Zhang, Yining Zheng, Xipeng Qiu</p><p><b>Upvotes:</b> 135</p><p><b>Summary:</b> Despite the success of reinforcement learning for large language models, a common failure mode is reduced sampling diversity, where the policy repeatedly generates similar erroneous behaviors. Classical entropy regularization encourages randomness under the current policy, but does not explicitly discourage recurrent failure patterns across rollouts. We propose MEDS, a Memory-Enhanced Dynamic reward Shaping framework that incorporates historical behavioral signals into reward design. By storing and leveraging intermediate model representations, we capture features of past rollouts and use density-based clustering to identify frequently recurring error patterns. Rollouts assigned to more prevalent error clusters are penalized more heavily, encouraging broader exploration while reducing repeated mistakes. Across five datasets and three base models, MEDS consistently improves average performance over existing baselines, achieving gains of up to 4.13 pass@1 points and 4.37 pass@128 points. Additional analyses using both LLM-based annotations and quantitative diversity metrics show that MEDS increases behavioral diversity during sampling.</p> + https://arxiv.org/abs/2604.11297 + Mon, 13 Apr 2026 00:00:00 +0000 + + + ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agents + https://arxiv.org/abs/2604.11784 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11784.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Fei Tang, Zhiqiong Lu, Boxuan Zhang, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen</p><p><b>Upvotes:</b> 137</p><p><b>Summary:</b> GUI agents drive applications through their visual interfaces instead of programmatic APIs, interacting with arbitrary software via taps, swipes, and keystrokes, reaching a long tail of applications that CLI-based agents cannot. Yet progress in this area is bottlenecked less by modeling capacity than by the absence of a coherent full-stack infrastructure: online RL training suffers from environment instability and closed pipelines, evaluation protocols drift silently across works, and trained agents rarely reach real users on real devices. We present ClawGUI, an open-source framework addressing these three gaps within a single harness. ClawGUI-RL provides the first open-source GUI agent RL infrastructure with validated support for both parallel virtual environments and real physical devices, integrating GiGPO with a Process Reward Model for dense step-level supervision. ClawGUI-Eval enforces a fully standardized evaluation pipeline across 6 benchmarks and 11+ models, achieving 95.8\% reproduction against official baselines. ClawGUI-Agent brings trained agents to Android, HarmonyOS, and iOS through 12+ chat platforms with hybrid CLI-GUI control and persistent personalized memory. Trained end to end within this pipeline, ClawGUI-2B achieves 17.1\% Success Rate on MobileWorld GUI-Only, outperforming the same-scale MAI-UI-2B baseline by 6.0\%.</p> + https://arxiv.org/abs/2604.11784 + Mon, 13 Apr 2026 00:00:00 +0000 + + + Self-Distillation Zero: Self-Revision Turns Binary Rewards into Dense Supervision + https://arxiv.org/abs/2604.12002 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12002.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yinghui He, Simran Kaur, Adithya Bhaskar, Yongjin Yang, Jiarui Liu, Narutatsu Ri, Liam Fowl, Abhishek Panigrahi, Danqi Chen, Sanjeev Arora</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Current post-training methods in verifiable settings fall into two categories. Reinforcement learning (RLVR) relies on binary rewards, which are broadly applicable and powerful, but provide only sparse supervision during training. Distillation provides dense token-level supervision, typically obtained from an external teacher or using high-quality demonstrations. Collecting such supervision can be costly or unavailable. We propose Self-Distillation Zero (SD-Zero), a method that is substantially more training sample-efficient than RL and does not require an external teacher or high-quality demonstrations. SD-Zero trains a single model to play two roles: a Generator, which produces an initial response, and a Reviser, which conditions on that response and its binary reward to produce an improved response. We then perform on-policy self-distillation to distill the reviser into the generator, using the reviser's token distributions conditioned on the generator's response and its reward as supervision. In effect, SD-Zero trains the model to transform binary rewards into dense token-level self-supervision. On math and code reasoning benchmarks with Qwen3-4B-Instruct and Olmo-3-7B-Instruct, SD-Zero improves performance by at least 10% over the base models and outperforms strong baselines, including Rejection Fine-Tuning (RFT), GRPO, and Self-Distillation Fine-Tuning (SDFT), under the same question set and training sample budget. Extensive ablation studies show two novel characteristics of our proposed algorithm: (a) token-level self-localization, where the reviser can identify the key tokens that need to be revised in the generator's response based on reward, and (b) iterative self-evolution, where the improving ability to revise answers can be distilled back into generation performance with regular teacher synchronization.</p> + https://arxiv.org/abs/2604.12002 + Mon, 13 Apr 2026 19:46:55 +0000 + + + Accelerating Speculative Decoding with Block Diffusion Draft Trees + https://arxiv.org/abs/2604.12989 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12989.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Liran Ringel, Yaniv Romano</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Speculative decoding accelerates autoregressive language models by using a lightweight drafter to propose multiple future tokens, which the target model then verifies in parallel. DFlash shows that a block diffusion drafter can generate an entire draft block in a single forward pass and achieve state-of-the-art speculative decoding performance, outperforming strong autoregressive drafters such as EAGLE-3. Vanilla DFlash, however, still verifies only a single drafted trajectory per round, potentially limiting its acceptance length. We introduce DDTree (Diffusion Draft Tree), a method that constructs a draft tree directly from the per-position distributions of a block diffusion drafter. Under a fixed node budget, DDTree uses a simple best-first heap algorithm to select the continuations that are most likely to match the target model according to a surrogate defined by the draft model's output. The resulting tree is verified efficiently in a single target model forward pass using an ancestor-only attention mask. Because DDTree builds on DFlash, a leading draft model for speculative decoding, these gains place DDTree among the leading approaches to speculative decoding.</p> + https://arxiv.org/abs/2604.12989 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Boosting Visual Instruction Tuning with Self-Supervised Guidance + https://arxiv.org/abs/2604.12966 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12966.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sophia Sirko-Galouchenko, Monika Wysoczanska, Andrei Bursuc, Nicolas Thome, Spyros Gidaris</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Multimodal large language models (MLLMs) perform well on many vision-language tasks but often struggle with vision-centric problems that require fine-grained visual reasoning. Recent evidence suggests that this limitation arises not from weak visual representations, but from under-utilization of visual information during instruction tuning, where many tasks can be partially solved using language priors alone. We propose a simple and lightweight approach that augments visual instruction tuning with a small number of visually grounded self-supervised tasks expressed as natural language instructions. By reformulating classical self-supervised pretext tasks, such as rotation prediction, color matching, and cross-view correspondence, as image-instruction-response triplets, we introduce supervision that cannot be solved without relying on visual evidence. Our approach requires no human annotations, no architectural modifications, and no additional training stages. Across multiple models, training regimes, and benchmarks, injecting only a small fraction (3-10%) of such visually grounded instructions consistently improves performance on vision-centric evaluations. Our findings highlight instruction tuning with visually grounded SSL tasks as a powerful lever for improving visual reasoning in MLLMs through simple adjustments to the training data distribution. Code available at: https://github.com/sirkosophia/V-GIFT</p> + https://arxiv.org/abs/2604.12966 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Masked by Consensus: Disentangling Privileged Knowledge in LLM Correctness + https://arxiv.org/abs/2604.12373 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12373.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tomer Ashuach, Liat Ein-Dor, Shai Gretz, Yoav Katz, Yonatan Belinkov</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Humans use introspection to evaluate their understanding through private internal states inaccessible to external observers. We investigate whether large language models possess similar privileged knowledge about answer correctness, information unavailable through external observation. We train correctness classifiers on question representations from both a model's own hidden states and external models, testing whether self-representations provide a performance advantage. On standard evaluation, we find no advantage: self-probes perform comparably to peer-model probes. We hypothesize this is due to high inter-model agreement of answer correctness. To isolate genuine privileged knowledge, we evaluate on disagreement subsets, where models produce conflicting predictions. Here, we discover domain-specific privileged knowledge: self-representations consistently outperform peer representations in factual knowledge tasks, but show no advantage in math reasoning. We further localize this domain asymmetry across model layers, finding that the factual advantage emerges progressively from early-to-mid layers onward, consistent with model-specific memory retrieval, while math reasoning shows no consistent advantage at any depth.</p> + https://arxiv.org/abs/2604.12373 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Grid2Matrix: Revealing Digital Agnosia in Vision-Language Models + https://arxiv.org/abs/2604.09687 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.09687.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yunkai Zhang, Linda Li, Yingxin Cui, Xiyuan Ruan, Zeyu Zheng, Kezhen Chen, Yi Zhang, Diji Yang</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Vision-Language Models (VLMs) excel on many multimodal reasoning benchmarks, but these evaluations often do not require an exhaustive readout of the image and can therefore obscure failures in faithfully capturing all visual details. We introduce Grid2Matrix (G2M), a controlled benchmark in which a model is shown a color grid and a color-to-number mapping, and must output the corresponding matrix. By varying grid size and the number of colors, G2M provides a simple way to increase visual complexity while minimizing semantic confounds. We find that VLMs exhibit a sharp early collapse in zero-shot end-to-end evaluation, failing on surprisingly small grids rather than degrading gradually as the task becomes denser. We probe the visual encoders of VLMs from two representative families and find that they preserve substantially more of the grid information than the corresponding end-to-end outputs. This suggests that the failure is not explained by visual encoding alone, but also reflects a gap between what remains recoverable from visual features and what is ultimately expressed in language. We term this gap Digital Agnosia. Further analyses show that these errors are highly structured and depend strongly on how grid cells overlap with visual patch boundaries. We also find that common strategies such as model scaling and multimodal alignment do not fully eliminate this failure mode. We expect G2M to serve as a useful testbed for understanding where and how VLMs lose fine visual details, and for evaluating tasks where missing even small visual details can matter, such as tables, charts, forms, and GUIs.</p> + https://arxiv.org/abs/2604.09687 + Tue, 14 Apr 2026 00:00:00 +0000 + + + KV Packet: Recomputation-Free Context-Independent KV Caching for LLMs + https://arxiv.org/abs/2604.13226 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13226.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chuangtao Chen, Grace Li Zhang, Xunzhao Yin, Cheng Zhuo, Bing Li, Ulf Schlichtmann</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Large Language Models (LLMs) rely heavily on Key-Value (KV) caching to minimize inference latency. However, standard KV caches are context-dependent: reusing a cached document in a new context requires recomputing KV states to account for shifts in attention distribution. Existing solutions such as CacheBlend, EPIC, and SAM-KV mitigate this issue by selectively recomputing a subset of tokens; however, they still incur non-negligible computational overhead (FLOPs) and increased Time-to-First-Token (TTFT) latency. In this paper, we propose KV Packet, a recomputation-free cache reuse framework that treats cached documents as immutable ``packets'' wrapped in light-weight trainable soft-token adapters, which are trained via self-supervised distillation to bridge context discontinuities. Experiments on Llama-3.1 and Qwen2.5 demonstrate that the proposed KV Packet method achieves near-zero FLOPs and lower TTFT than recomputation-based baselines, while retaining F1 scores comparable to those of the full recomputation baseline.</p> + https://arxiv.org/abs/2604.13226 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Lightning OPD: Efficient Post-Training for Large Reasoning Models with Offline On-Policy Distillation + https://arxiv.org/abs/2604.13010 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13010.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yecheng Wu, Song Han, Hai Cai</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> On-policy distillation (OPD) has emerged as an efficient post-training paradigm for large language models. However, standard OPD requires a live teacher inference server throughout training, resulting in substantial infrastructure overhead. In this work, we investigate whether on-policy distillation can be performed offline. A natural approach is to precompute teacher log-probabilities once over SFT rollouts and reuse them during training. In practice, however, this offline variant fails to reliably match the performance of standard OPD. To understand this discrepancy, we identify a previously overlooked condition that is critical for any OPD pipeline, which we term teacher consistency. This condition requires that the same teacher model be used for both supervised fine-tuning and OPD. We show that violating teacher consistency introduces an irreducible gradient bias, causing both offline and online OPD to converge to a suboptimal fixed point regardless of training duration. Building on this insight, we propose Lightning OPD, an offline on-policy distillation framework that enforces teacher consistency by precomputing teacher log-probabilities over SFT rollouts. This design eliminates the need for a live teacher server entirely. We further show that, under teacher consistency, Lightning OPD shares the same optimum as standard OPD, with bounded gradient discrepancy and an implicit regularization effect that helps prevent policy drift. Extensive experiments on mathematical reasoning and code generation demonstrate that Lightning OPD achieves state-of-the-art performance with significantly improved efficiency. Starting from an SFT-initialized Qwen3-8B-Base model, Lightning OPD reaches 69.9% on AIME 2024 in just 30 GPU hours, achieving a 4.0x speedup over standard OPD and substantially lowering the barrier to entry for academic research on LLM post-training.</p> + https://arxiv.org/abs/2604.13010 + Tue, 14 Apr 2026 00:00:00 +0000 + + + ROSE: An Intent-Centered Evaluation Metric for NL2SQL + https://arxiv.org/abs/2604.12988 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12988.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wenqi Pei, Shizheng Hou, Boyan Li, Han Chen, Zhichao Shi, Yuyu Luo</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> Execution Accuracy (EX), the widely used metric for evaluating the effectiveness of Natural Language to SQL (NL2SQL) solutions, is becoming increasingly unreliable. It is sensitive to syntactic variation, ignores that questions may admit multiple interpretations, and is easily misled by erroneous ground-truth SQL. To address this, we introduce ROSE, an intent-centered metric that focuses on whether the predicted SQL answers the question, rather than consistency with the ground-truth SQL under the reference-dependent paradigm. ROSE employs an adversarial Prover-Refuter cascade: SQL Prover assesses the semantic correctness of a predicted SQL against the user's intent independently, while Adversarial Refuter uses the ground-truth SQL as evidence to challenge and refine this judgment. On our expert-aligned validation set ROSE-VEC, ROSE achieves the best agreement with human experts, outperforming the next-best metric by nearly 24% in Cohen's Kappa. We also conduct a largescale re-evaluation of 19 NL2SQL methods, revealing four valuable insights. We release ROSE and ROSE-VEC to facilitate more reliable NL2SQL research.</p> + https://arxiv.org/abs/2604.12988 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Self-Adversarial One Step Generation via Condition Shifting + https://arxiv.org/abs/2604.12322 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12322.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Deyuan Liu, Peng Sun, Yansen Han, Zhenglin Cheng, Chuyan Chen, Tao Lin</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> The push for efficient text to image synthesis has moved the field toward one step sampling, yet existing methods still face a three way tradeoff among fidelity, inference speed, and training efficiency. Approaches that rely on external discriminators can sharpen one step performance, but they often introduce training instability, high GPU memory overhead, and slow convergence, which complicates scaling and parameter efficient tuning. In contrast, regression based distillation and consistency objectives are easier to optimize, but they typically lose fine details when constrained to a single step. We present APEX, built on a key theoretical insight: adversarial correction signals can be extracted endogenously from a flow model through condition shifting. Using a transformation creates a shifted condition branch whose velocity field serves as an independent estimator of the model's current generation distribution, yielding a gradient that is provably GAN aligned, replacing the sample dependent discriminator terms that cause gradient vanishing. This discriminator free design is architecture preserving, making APEX a plug and play framework compatible with both full parameter and LoRA based tuning. Empirically, our 0.6B model surpasses FLUX-Schnell 12B (20times more parameters) in one step quality. With LoRA tuning on Qwen-Image 20B, APEX reaches a GenEval score of 0.89 at NFE=1 in 6 hours, surpassing the original 50-step teacher (0.87) and providing a 15.33times inference speedup. Code is available https://github.com/LINs-lab/APEX.</p> + https://arxiv.org/abs/2604.12322 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Habitat-GS: A High-Fidelity Navigation Simulator with Dynamic Gaussian Splatting + https://arxiv.org/abs/2604.12626 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12626.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ziyuan Xia, Jingyi Xu, Chong Cui, Yuanhong Yu, Jiazhao Zhang, Qingsong Yan, Tao Ni, Junbo Chen, Xiaowei Zhou, Hujun Bao, Ruizhen Hu, Sida Peng</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> Training embodied AI agents depends critically on the visual fidelity of simulation environments and the ability to model dynamic humans. Current simulators rely on mesh-based rasterization with limited visual realism, and their support for dynamic human avatars, where available, is constrained to mesh representations, hindering agent generalization to human-populated real-world scenarios. We present Habitat-GS, a navigation-centric embodied AI simulator extended from Habitat-Sim that integrates 3D Gaussian Splatting scene rendering and drivable gaussian avatars while maintaining full compatibility with the Habitat ecosystem. Our system implements a 3DGS renderer for real-time photorealistic rendering and supports scalable 3DGS asset import from diverse sources. For dynamic human modeling, we introduce a gaussian avatar module that enables each avatar to simultaneously serve as a photorealistic visual entity and an effective navigation obstacle, allowing agents to learn human-aware behaviors in realistic settings. Experiments on point-goal navigation demonstrate that agents trained on 3DGS scenes achieve stronger cross-domain generalization, with mixed-domain training being the most effective strategy. Evaluations on avatar-aware navigation further confirm that gaussian avatars enable effective human-aware navigation. Finally, performance benchmarks validate the system's scalability across varying scene complexity and avatar counts.</p> + https://arxiv.org/abs/2604.12626 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Generative Refinement Networks for Visual Synthesis + https://arxiv.org/abs/2604.13030 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13030.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jian Han, Jinlai Liu, Jiahuan Wang, Bingyue Peng, Zehuan Yuan</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> While diffusion models dominate the field of visual generation, they are computationally inefficient, applying a uniform computational effort regardless of different complexity. In contrast, autoregressive (AR) models are inherently complexity-aware, as evidenced by their variable likelihoods, but are often hindered by lossy discrete tokenization and error accumulation. In this work, we introduce Generative Refinement Networks (GRN), a next-generation visual synthesis paradigm to address these issues. At its core, GRN addresses the discrete tokenization bottleneck through a theoretically near-lossless Hierarchical Binary Quantization (HBQ), achieving a reconstruction quality comparable to continuous counterparts. Built upon HBQ's latent space, GRN fundamentally upgrades AR generation with a global refinement mechanism that progressively perfects and corrects artworks -- like a human artist painting. Besides, GRN integrates an entropy-guided sampling strategy, enabling complexity-aware, adaptive-step generation without compromising visual quality. On the ImageNet benchmark, GRN establishes new records in image reconstruction (0.56 rFID) and class-conditional image generation (1.81 gFID). We also scale GRN to more challenging text-to-image and text-to-video generation, delivering superior performance on an equivalent scale. We release all models and code to foster further research on GRN.</p> + https://arxiv.org/abs/2604.13030 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Dive into Claude Code: The Design Space of Today's and Future AI Agent Systems + https://arxiv.org/abs/2604.14228 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14228.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiacheng Liu, Xiaohan Zhao, Xinyi Shang, Zhiqiang Shen</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> Claude Code is an agentic coding tool that can run shell commands, edit files, and call external services on behalf of the user. This study describes its comprehensive architecture by analyzing the publicly available TypeScript source code and further comparing it with OpenClaw, an independent open-source AI agent system that answers many of the same design questions from a different deployment context. Our analysis identifies five human values, philosophies, and needs that motivate the architecture (human decision authority, safety and security, reliable execution, capability amplification, and contextual adaptability) and traces them through thirteen design principles to specific implementation choices. The core of the system is a simple while-loop that calls the model, runs tools, and repeats. Most of the code, however, lives in the systems around this loop: a permission system with seven modes and an ML-based classifier, a five-layer compaction pipeline for context management, four extensibility mechanisms (MCP, plugins, skills, and hooks), a subagent delegation mechanism with worktree isolation, and append-oriented session storage. A comparison with OpenClaw, a multi-channel personal assistant gateway, shows that the same recurring design questions produce different architectural answers when the deployment context changes: from per-action safety classification to perimeter-level access control, from a single CLI loop to an embedded runtime within a gateway control plane, and from context-window extensions to gateway-wide capability registration. We finally identify six open design directions for future agent systems, grounded in recent empirical, architectural, and policy literature.</p> + https://arxiv.org/abs/2604.14228 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Towards Long-horizon Agentic Multimodal Search + https://arxiv.org/abs/2604.12890 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12890.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yifan Du, Zikang Liu, Jinbiao Peng, Jie Wu, Junyi Li, Jinyang Li, Wayne Xin Zhao, Ji-Rong Wen</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Multimodal deep search agents have shown great potential in solving complex tasks by iteratively collecting textual and visual evidence. However, managing the heterogeneous information and high token costs associated with multimodal inputs over long horizons remains a critical challenge, as existing methods often suffer from context explosion or the loss of crucial visual signals. To address this, we propose a novel Long-horizon MultiModal deep search framework, named LMM-Searcher, centered on a file-based visual representation mechanism. By offloading visual assets to an external file system and mapping them to lightweight textual identifiers (UIDs), our approach mitigates context overhead while preserving multimodal information for future access. We equip the agent with a tailored fetch-image tool, enabling a progressive, on-demand visual loading strategy for active perception. Furthermore, we introduce a data synthesis pipeline designed to generate queries requiring complex cross-modal multi-hop reasoning. Using this pipeline, we distill 12K high-quality trajectories to fine-tune Qwen3-VL-Thinking-30A3B into a specialized multimodal deep search agent. Extensive experiments across four benchmarks demonstrate that our method successfully scales to 100-turn search horizons, achieving state-of-the-art performance among open-source models on challenging long-horizon benchmarks like MM-BrowseComp and MMSearch-Plus, while also exhibiting strong generalizability across different base models. Our code will be released in https://github.com/RUCAIBox/LMM-Searcher.</p> + https://arxiv.org/abs/2604.12890 + Tue, 14 Apr 2026 00:00:00 +0000 + + + ASGuard: Activation-Scaling Guard to Mitigate Targeted Jailbreaking Attack + https://arxiv.org/abs/2509.25843 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2509.25843.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yein Park, Jungwoo Park, Jaewoo Kang</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Large language models (LLMs), despite being safety-aligned, exhibit brittle refusal behaviors that can be circumvented by simple linguistic changes. As tense jailbreaking demonstrates that models refusing harmful requests often comply when rephrased in past tense, a critical generalization gap is revealed in current alignment methods whose underlying mechanisms are poorly understood. In this work, we introduce Activation-Scaling Guard (ASGuard), an insightful, mechanistically-informed framework that surgically mitigates this specific vulnerability. In the first step, we use circuit analysis to identify the specific attention heads causally linked to the targeted jailbreaking such as a tense-changing attack. Second, we train a precise, channel-wise scaling vector to recalibrate the activation of tense vulnerable heads. Lastly, we apply it into a "preventative fine-tuning", forcing the model to learn a more robust refusal mechanism. Across four LLMs, ASGuard effectively reduces the attack success rate of targeted jailbreaking while preserving general capabilities and minimizing over refusal, achieving a Pareto-optimal balance between safety and utility. Our findings underscore how adversarial suffixes suppress the propagation of the refusal-mediating direction, based on mechanistic analysis. Furthermore, our work showcases how a deep understanding of model internals can be leveraged to develop practical, efficient, and targeted methods for adjusting model behavior, charting a course for more reliable and interpretable AI safety.</p> + https://arxiv.org/abs/2509.25843 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Exploration and Exploitation Errors Are Measurable for Language Model Agents + https://arxiv.org/abs/2604.13151 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13151.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jaden Park, Jungtaek Kim, Jongwon Jeong, Robert D. Nowak, Kangwook Lee, Yong Jae Lee</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> Language Model (LM) agents are increasingly used in complex open-ended decision-making tasks, from AI coding to physical AI. A core requirement in these settings is the ability to both explore the problem space and exploit acquired knowledge effectively. However, systematically distinguishing and quantifying exploration and exploitation from observed actions without access to the agent's internal policy remains challenging. To address this, we design controllable environments inspired by practical embodied AI scenarios. Each environment consists of a partially observable 2D grid map and an unknown task Directed Acyclic Graph (DAG). The map generation can be programmatically adjusted to emphasize exploration or exploitation difficulty. To enable policy-agnostic evaluation, we design a metric to quantify exploration and exploitation errors from agent's actions. We evaluate a variety of frontier LM agents and find that even state-of-the-art models struggle on our task, with different models exhibiting distinct failure modes. We further observe that reasoning models solve the task more effectively and show both exploration and exploitation can be significantly improved through minimal harness engineering. We release our code https://github.com/jjj-madison/measurable-explore-exploit{here}.</p> + https://arxiv.org/abs/2604.13151 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning + https://arxiv.org/abs/2604.12374 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12374.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> NVIDIA, Aakshita Chandiramani, Aaron Blakeman, Abdullahi Olaoye, Abhibha Gupta, Abhilash Somasamudramath, Abhinav Khattar, Adeola Adesoba, Adi Renduchintala, Adil Asif, Aditya Agrawal, Aditya Vavre, Ahmad Kiswani, Aishwarya Padmakumar, Ajay Hotchandani, Akanksha Shukla, Akhiad Bercovich, Aleksander Ficek, Aleksandr Shaposhnikov, Alex Gronskiy, Alex Kondratenko, Alex Neefus, Alex Steiner, Alex Yang, Alexander Bukharin, Alexander Young, Ali Hatamizadeh, Ali Taghibakhshi, Alina Galiautdinova, Alisa Liu, Alok Kumar, Ameya Sunil Mahabaleshwarkar, Amir Klein, Amit Zuker, Amnon Geifman, Anahita Bhiwandiwalla, Ananth Subramaniam, Andrew Tao, Anjaney Shrivastava, Anjulie Agrusa, Ankur Srivastava, Ankur Verma, Ann Guan, Anna Shors, Annamalai Chockalingam, Anubhav Mandarwal, Aparnaa Ramani, Arham Mehta, Arti Jain, Arun Venkatesan, Asha Anoosheh, Ashwath Aithal, Ashwin Poojary, Asif Ahamed, Asit Mishra, Asli Sabanci Demiroz, Asma Kuriparambil Thekkumpate, Atefeh Sohrabizadeh, Avinash Kaur, Ayush Dattagupta, Barath Subramaniam Anandan, Bardiya Sadeghi, Barnaby Simkin, Ben Lanir, Benedikt Schifferer, Benjamin Chislett, Besmira Nushi, Bilal Kartal, Bill Thiede, Bita Darvish Rouhani, Bobby Chen, Boris Ginsburg, Brandon Norick, Branislav Kisacanin, Brian Yu, Bryan Catanzaro, Buvaneswari Mani, Carlo del Mundo, Chankyu Lee, Chanran Kim, Chantal Hwang, Chao Ni, Charles Wang, Charlie Truong, Cheng-Ping Hsieh, Chenhan Yu, Chenjie Luo, Cherie Wang, Chetan Mungekar, Chintan Patel, Chris Alexiuk, Chris Holguin, Chris Wing, Christian Munley, Christopher Parisien, Chuck Desai, Chunyang Sheng, Collin Neale, Cyril Meurillon, Dakshi Kumar, Dan Gil, Dan Su, Dane Corneil, Daniel Afrimi, Daniel Burkhardt Eliuth Triana, Daniel Egert, Daniel Fatade, Daniel Lo, Daniel Rohrer, Daniel Serebrenik, Daniil Sorokin, Daria Gitman, Daria Levy, Darko Stosic, David Edelsohn, David Messina, David Mosallanezhad, David Tamok, Deena Donia, Deepak Narayanan, Devin O'Kelly, Dheeraj Peri, Dhruv Nathawani, Di Wu, Dima Rekesh, Dina Yared, Divyanshu Kakwani, Dmitry Konyagin Brandon Tuttle, Dong Ahn, Dongfu Jiang, Dorrin Poorkay, Douglas O'Flaherty, Duncan Riach, Dusan Stosic, Dustin Van Stee, Edgar Minasyan, Edward Lin, Eileen Peters Long, Elad Segal, Elena Lantz, Elena Lewis, Ellie Evans, Elliott Ning, Eric Chung, Eric Harper, Eric Pham-Hung, Eric W. Tramel, Erick Galinkin, Erik Pounds, Esti Etrog, Evan Briones, Evan Wu, Evelina Bakhturina, Evgeny Tsykunov, Ewa Dobrowolska, Farshad Saberi Movahed, Farzan Memarian, Fay Wang, Fei Jia, Felipe Soares, Felipe Vieira Frujeri, Feng Chen, Fengguang Lin, Ferenc Galko, Fortuna Zhang, Frankie Siino, Frida Hou, Gantavya Bhatt, Gargi Prasad, Geethapriya Venkataramani, Geetika Gupta, George Armstrong, Gerald Shen, Giulio Borghesi, Gordana Neskovic, Gorkem Batmaz, Grace Lam, Grace Wu, Greg Pauloski, Greyson Davis, Grigor Nalbandyan, Guoming Zhang, Guy Farber, Guyue Huang, Haifeng Qian, Haran Kumar Shiv Kumar, Harry Kim, Harsh Sharma, Hayate Iso, Hayley Ross, Herbert Hum, Herman Sahota, Hexin Wang, Himanshu Soni, Hiren Upadhyay, Huy Nguyen, Iain Cunningham, Ido Galil, Ido Shahaf, Igino Padovani, Igor Gitman, Igor Shovkun, Ikroop Dhillon, Ilya Loshchilov, Ingrid Kelly, Itamar Schen, Itay Levy, Ivan Moshkov, Izik Golan, Izzy Putterman, Jain Tu, Jan Baczek, Jan Kautz, Jane Polak Scowcroft, Janica Rosenberg, Jared Casper, Jarrod Pflum, Jason Grant, Jason Sewall, Jatin Mitra, Jeffrey Glick, Jenny Chen, Jesse Oliver, Jiacheng Xu, Jiafan Zhu, Jialin Song, Jian Zhang, Jiaqi Zeng, Jie Lou, Jill Milton, Jim Chow, Jimmy Zhang, Jinhang Choi, Jining Huang, Jocelyn Huang, Joel Caruso, Joey Conway, Joey Guman, Johan Jatko, John Kamalu, Johnny Greco, Jonathan Cohen, Jonathan Raiman, Joseph Jennings, Joyjit Daw, Juan Yu, Julio Tapia, Junkeun Yi, Jupinder Parmar, Jyothi Achar, Kari Briski, Kartik Mattoo, Katherine Cheung, Katherine Luna, Keith Wyss, Kevin Shih, Kezhi Kong, Khanh Nguyen, Khushi Bhardwaj, Kirill Buryak, Kirthi Shankar Sivamani, Konstantinos Krommydas, Kris Murphy, Krishna C. Puvvada, Krzysztof Pawelec, Kumar Anik, Laikh Tewari, Laya Sleiman, Leo Du, Leon Derczynski, Li Ding, Lilach Ilan, Lingjie Wu, Lizzie Wei, Luis Vega, Lun Su, Maarten Van Segbroeck, Maer Rodrigues de Melo, Magaret Zhang, Mahan Fathi, Makesh Narsimhan Sreedhar, Makesh Sreedhar, Makesh Tarun Chandran, Manuel Reyes Gomez, Maor Ashkenazi, Marc Cuevas, Marc Romeijn, Margaret Zhang, Mark Cai, Mark Gabel, Markus Kliegl, Martyna Patelka, Maryam Moosaei, Matthew Varacalli, Matvei Novikov, Mauricio Ferrato, Mehrzad Samadi, Melissa Corpuz, Meng Xin, Mengdi Wang, Mengru Wang, Meredith Price, Micah Schaffer, Michael Andersch, Michael Boone, Michael Evans, Michael Z Wang, Miguel Martinez, Mikail Khona, Mike Chrzanowski, Mike Hollinger, Mingyuan Ma, Minseok Lee, Mohammad Dabbah, Mohammad Shoeybi, Mostofa Patwary, Nabin Mulepati, Nader Khalil, Najeeb Nabwani, Nancy Agarwal, Nanthini Balasubramaniam, Narimane Hennouni, Narsi Kodukula, Natalie Hereth, Nathaniel Pinckney, Nave Assaf, Negar Habibi, Nestor Qin, Neta Zmora, Netanel Haber, Nick Reamaroon, Nickson Quak, Nidhi Bhatia, Nikhil Jukar, Nikki Pope, Nikolai Ludwig, Nima Tajbakhsh, Nir Ailon, Nirmal Juluru, Nirmalya De, Nowel Pitt, Oleg Rybakov, Oleksii Hrinchuk, Oleksii Kuchaiev, Olivier Delalleau, Oluwatobi Olabiyi, Omer Ullman Argov, Omri Almog, Omri Puny, Oren Tropp, Otavio Padovani, Ouye Xie, Parth Chadha, Pasha Shamis, Paul Gibbons, Pavlo Molchanov, Peter Belcak, Peter Jin, Pinky Xu, Piotr Januszewski, Pooya Jannaty, Prachi Shevate, Pradeep Thalasta, Pranav Prashant Thombre, Prasoon Varshney, Prerana Gambhir, Pritam Gundecha, Przemek Tredak, Qing Miao, Qiyu Wan, Quan Tran Minh, Rabeeh Karimi Mahabadi, Rachel Oberman, Rachit Garg, Rahul Kandu, Raina Zhong, Ran El-Yaniv, Ran Zilberstein, Rasoul Shafipour, Renee Yao, Renjie Pi, Richard Mazzarese, Richard Wang, Rick Izzo, Ridhima Singla, Rima Shahbazyan, Rishabh Garg, Ritika Borkar, Ritu Gala, Riyad Islam, Robert Clark, Robert Hesse, Roger Waleffe, Rohit Varma Kalidindi, Rohit Watve, Roi Koren, Ron Fan, Ruchika Kharwar, Ruisi Cai, Ruoxi Zhang, Russell J. Hewett, Ryan Prenger, Ryan Timbrook, Ryota Egashira, Sadegh Mahdavi, Sagar Singh Ashutosh Joshi, Sahil Modi, Samuel Kriman, Sandeep Pombra, Sanjay Kariyappa, Sanjeev Satheesh, Santiago Pombo, Saori Kaji, Satish Pasumarthi, Saurav Mishra, Saurav Muralidharan, Scott Hara, Sean Narenthiran, Sebastian Rogawski, Seonjin Na, Seonmyeong Bak, Sepehr Sameni, Seth Poulos, Shahar Mor, Shantanu Acharya, Shaona Ghosh Adam Lord, Sharath Turuvekere Sreenivas, Shaun Kotek, Shaya Gharghabi, Shelby Thomas, Sheng-Chieh Lin, Shibani Likhite, Shiqing Fan, Shiyang Chen, Shreya Gopal, Shrimai Prabhumoye, Shubham Pachori, Shubham Toshniwal, Shuo Zhang, Shuoyang Ding, Shyam Renjith, Shyamala Prayaga, Siddhartha Jain, Simeng Sun, Sirisha Rella, Sirshak Das, Smita Ithape, Sneha Harishchandra S, Somshubra Majumdar, Soumye Singhal, Sri Harsha Singudasu, Sriharsha Niverty, Stas Sergienko, Stefana Gloginic, Stefania Alborghetti, Stephen Ge, Stephen McCullough, Sugam Dipak Devare, Suguna Varshini Velury, Sukrit Rao, Sumeet Kumar Barua, Sunny Gai, Suseella Panguluri, Sushil Koundinyan, Swathi Patnam, Sweta Priyadarshi, Swetha Bhendigeri, Syeda Nahida Akter, Sylendran Arunagiri, Tailling Yuan, Talor Abramovich, Tan Bui, Tan Yu, Terry Kong, Thanh Do, Thomas Gburek, Thorgane Marques, Tiffany Moore, Tijmen Blankevoort, Tim Moon, Timothy Ma, Tiyasa Mitra, Tomasz Grzegorzek, Tomer Asida, Tomer Bar Natan, Tomer Keren, Tomer Ronen, Traian Rebedea, Trenton Starkey, Tugrul Konuk, Twinkle Vashishth, Tyler Condensa, Udi Karpas, Ushnish De, Vahid Noorozi, Vahid Noroozi, Vanshil Atul Shah, Veena Vaidyanathan, Venkat Srinivasan, Venmugil Elango, Victor Cui, Vijay Korthikanti, Vikas Mehta, Virginia Adams, Virginia Wu, Vitaly Kurin, Vitaly Lavrukhin, Vladimir Anisimov, Wan Seo, Wanli Jiang, Wasi Uddin Ahmad, Wei Du, Wei Ping, Wei-Ming Chen, Wendy Quan, Wenliang Dai, Wenwen Gao, Will Jennings, William Zhang, Xiaowei Ren, Xiaowen Xin, Xin Li, Yang Yu, Yangyi Chen, Yaniv Galron, Yashaswi Karnati, Yejin Choi, Yev Meyer, Yi-Fu Wu, Yian Zhang, Ying Lin, Yonatan Geifman, Yonggan Fu, Yoshi Suhara, Youngeun Kwon, Yuan Zhang, Yuki Huang, Zach Moshe, Zhilin Wang, Zhiyu Cheng, Zhongbo Zhu, Zhuolin Yang, Zihan Liu, Zijia Chen, Zijie Yan, Zuhair Ahmed</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> We describe the pre-training, post-training, and quantization of Nemotron 3 Super, a 120 billion (active 12 billion) parameter hybrid Mamba-Attention Mixture-of-Experts model. Nemotron 3 Super is the first model in the Nemotron 3 family to 1) be pre-trained in NVFP4, 2) leverage LatentMoE, a new Mixture-of-Experts architecture that optimizes for both accuracy per FLOP and accuracy per parameter, and 3) include MTP layers for inference acceleration through native speculative decoding. We pre-trained Nemotron 3 Super on 25 trillion tokens followed by post-training using supervised fine tuning (SFT) and reinforcement learning (RL). The final model supports up to 1M context length and achieves comparable accuracy on common benchmarks, while also achieving up to 2.2x and 7.5x higher inference throughput compared to GPT-OSS-120B and Qwen3.5-122B, respectively. Nemotron 3 Super datasets, along with the base, post-trained, and quantized checkpoints, are open-sourced on HuggingFace.</p> + https://arxiv.org/abs/2604.12374 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Toward Autonomous Long-Horizon Engineering for ML Research + https://arxiv.org/abs/2604.13018 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13018.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Guoxin Chen, Jie Chen, Lei Chen, Jiale Zhao, Fanzhe Meng, Wayne Xin Zhao, Ruihua Song, Cheng Chen, Ji-Rong Wen, Kai Jia</p><p><b>Upvotes:</b> 34</p><p><b>Summary:</b> Autonomous AI research has advanced rapidly, but long-horizon ML research engineering remains difficult: agents must sustain coherent progress across task comprehension, environment setup, implementation, experimentation, and debugging over hours or days. We introduce AiScientist, a system for autonomous long-horizon engineering for ML research built on a simple principle: strong long-horizon performance requires both structured orchestration and durable state continuity. To this end, AiScientist combines hierarchical orchestration with a permission-scoped File-as-Bus workspace: a top-level Orchestrator maintains stage-level control through concise summaries and a workspace map, while specialized agents repeatedly re-ground on durable artifacts such as analyses, plans, code, and experimental evidence rather than relying primarily on conversational handoffs, yielding thin control over thick state. Across two complementary benchmarks, AiScientist improves PaperBench score by 10.54 points on average over the best matched baseline and achieves 81.82 Any Medal% on MLE-Bench Lite. Ablation studies further show that File-as-Bus protocol is a key driver of performance, reducing PaperBench by 6.41 points and MLE-Bench Lite by 31.82 points when removed. These results suggest that long-horizon ML research engineering is a systems problem of coordinating specialized work over durable project state, rather than a purely local reasoning problem.</p> + https://arxiv.org/abs/2604.13018 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Lyra 2.0: Explorable Generative 3D Worlds + https://arxiv.org/abs/2604.13036 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13036.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tianchang Shen, Sherwin Bahmani, Kai He, Sangeetha Grama Srinivasan, Tianshi Cao, Jiawei Ren, Ruilong Li, Zian Wang, Nicholas Sharp, Zan Gojcic, Sanja Fidler, Jiahui Huang, Huan Ling, Jun Gao, Xuanchi Ren</p><p><b>Upvotes:</b> 36</p><p><b>Summary:</b> Recent advances in video generation enable a new paradigm for 3D scene creation: generating camera-controlled videos that simulate scene walkthroughs, then lifting them to 3D via feed-forward reconstruction techniques. This generative reconstruction approach combines the visual fidelity and creative capacity of video models with 3D outputs ready for real-time rendering and simulation. Scaling to large, complex environments requires 3D-consistent video generation over long camera trajectories with large viewpoint changes and location revisits, a setting where current video models degrade quickly. Existing methods for long-horizon generation are fundamentally limited by two forms of degradation: spatial forgetting and temporal drifting. As exploration proceeds, previously observed regions fall outside the model's temporal context, forcing the model to hallucinate structures when revisited. Meanwhile, autoregressive generation accumulates small synthesis errors over time, gradually distorting scene appearance and geometry. We present Lyra 2.0, a framework for generating persistent, explorable 3D worlds at scale. To address spatial forgetting, we maintain per-frame 3D geometry and use it solely for information routing -- retrieving relevant past frames and establishing dense correspondences with the target viewpoints -- while relying on the generative prior for appearance synthesis. To address temporal drifting, we train with self-augmented histories that expose the model to its own degraded outputs, teaching it to correct drift rather than propagate it. Together, these enable substantially longer and 3D-consistent video trajectories, which we leverage to fine-tune feed-forward reconstruction models that reliably recover high-quality 3D scenes.</p> + https://arxiv.org/abs/2604.13036 + Tue, 14 Apr 2026 00:00:00 +0000 + + + Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe + https://arxiv.org/abs/2604.13016 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13016.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yaxuan Li, Yuxin Zuo, Bingxiang He, Jinqian Zhang, Chaojun Xiao, Cheng Qian, Tianyu Yu, Huan-ang Gao, Wenkai Yang, Zhiyuan Liu, Ning Ding</p><p><b>Upvotes:</b> 79</p><p><b>Summary:</b> On-policy distillation (OPD) has become a core technique in the post-training of large language models, yet its training dynamics remain poorly understood. This paper provides a systematic investigation of OPD dynamics and mechanisms. We first identify that two conditions govern whether OPD succeeds or fails: (i) the student and teacher should share compatible thinking patterns; and (ii) even with consistent thinking patterns and higher scores, the teacher must offer genuinely new capabilities beyond what the student has seen during training. We validate these findings through weak-to-strong reverse distillation, showing that same-family 1.5B and 7B teachers are distributionally indistinguishable from the student's perspective. Probing into the token-level mechanism, we show that successful OPD is characterized by progressive alignment on high-probability tokens at student-visited states, a small shared token set that concentrates most of the probability mass (97%-99%). We further propose two practical strategies to recover failing OPD: off-policy cold start and teacher-aligned prompt selection. Finally, we show that OPD's apparent free lunch of dense token-level reward comes at a cost, raising the question of whether OPD can scale to long-horizon distillation.</p> + https://arxiv.org/abs/2604.13016 + Tue, 14 Apr 2026 00:00:00 +0000 + + + KnowRL: Boosting LLM Reasoning via Reinforcement Learning with Minimal-Sufficient Knowledge Guidance + https://arxiv.org/abs/2604.12627 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.12627.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Linhao Yu, Tianmeng Yang, Siyu Ding, Renren Jin, Naibin Gu, Xiangzhao Hao, Shuaiyi Nie, Deyi Xiong, Weichong Yin, Yu Sun, Hua Wu</p><p><b>Upvotes:</b> 96</p><p><b>Summary:</b> RLVR improves reasoning in large language models, but its effectiveness is often limited by severe reward sparsity on hard problems. Recent hint-based RL methods mitigate sparsity by injecting partial solutions or abstract templates, yet they typically scale guidance by adding more tokens, which introduce redundancy, inconsistency, and extra training overhead. We propose KnowRL (Knowledge-Guided Reinforcement Learning), an RL training framework that treats hint design as a minimal-sufficient guidance problem. During RL training, KnowRL decomposes guidance into atomic knowledge points (KPs) and uses Constrained Subset Search (CSS) to construct compact, interaction-aware subsets for training. We further identify a pruning interaction paradox -- removing one KP may help while removing multiple such KPs can hurt -- and explicitly optimize for robust subset curation under this dependency structure. We train KnowRL-Nemotron-1.5B from OpenMath-Nemotron-1.5B. Across eight reasoning benchmarks at the 1.5B scale, KnowRL-Nemotron-1.5B consistently outperforms strong RL and hinting baselines. Without KP hints at inference, KnowRL-Nemotron-1.5B reaches 70.08 average accuracy, already surpassing Nemotron-1.5B by +9.63 points; with selected KPs, performance improves to 74.16, establishing a new state of the art at this scale. The model, curated training data, and code are publicly available at https://github.com/Hasuer/KnowRL.</p> + https://arxiv.org/abs/2604.12627 + Tue, 14 Apr 2026 00:00:00 +0000 + + + UI-Copilot: Advancing Long-Horizon GUI Automation via Tool-Integrated Policy Optimization + https://arxiv.org/abs/2604.13822 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13822.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhengxi Lu, Fei Tang, Guangyi Liu, Kaitao Song, Xu Tan, Jin Ma, Wenqi Zhang, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> MLLM-based GUI agents have demonstrated strong capabilities in complex user interface interaction tasks. However, long-horizon scenarios remain challenging, as these agents are burdened with tasks beyond their intrinsic capabilities, suffering from memory degradation, progress confusion, and math hallucination. To address these challenges, we present UI-Copilot, a collaborative framework where the GUI agent focuses on task execution while a lightweight copilot provides on-demand assistance for memory retrieval and numerical computation. We introduce memory decoupling to separate persistent observations from transient execution context, and train the policy agent to selectively invoke the copilot as Retriever or Calculator based on task demands. To enable effective tool invocation learning, we propose Tool-Integrated Policy Optimization (TIPO), which separately optimizes tool selection through single-turn prediction and task execution through on-policy multi-turn rollouts. Experimental results show that UI-Copilot-7B achieves state-of-the-art performance on challenging MemGUI-Bench, outperforming strong 7B-scale GUI agents such as GUI-Owl-7B and UI-TARS-1.5-7B. Moreover, UI-Copilot-7B delivers a 17.1% absolute improvement on AndroidWorld over the base Qwen model, highlighting UI-Copilot's strong generalization to real-world GUI tasks.</p> + https://arxiv.org/abs/2604.13822 + Wed, 15 Apr 2026 00:00:00 +0000 + + + MERRIN: A Benchmark for Multimodal Evidence Retrieval and Reasoning in Noisy Web Environments + https://arxiv.org/abs/2604.13418 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.13418.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Han Wang, David Wan, Hyunji Lee, Thinh Pham, Mikaela Cankosyan, Weiyuan Chen, Elias Stengel-Eskin, Tu Vu, Mohit Bansal</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Motivated by the underspecified, multi-hop nature of search queries and the multimodal, heterogeneous, and often conflicting nature of real-world web results, we introduce MERRIN (Multimodal Evidence Retrieval and Reasoning in Noisy Web Environments), a human-annotated benchmark for evaluating search-augmented agents. MERRIN measures AI agents' ability to identify relevant modalities, retrieve multimodal evidence, and perform multi-hop reasoning over noisy web sources. It differs from prior work in three important aspects: (1) using natural language queries without explicit modality cues, (2) incorporating underexplored modalities such as video and audio, and (3) requiring the retrieval of complex, often noisy or conflicting multimodal evidence during web search. We evaluate diverse search agents powered by ten models, including strong closed-source models (e.g., GPT-5.4-mini, Gemini 3/3.1 Flash/Pro) and open-weight models (Qwen3-4B/30B/235B), across three search settings (no search, native search, and agentic search). Our results show that MERRIN is highly challenging: the average accuracy across all agents is 22.3%, with the best-performing agent reaching only 40.1%. We further observe that while stronger agents like Gemini Deep Research achieve higher performance, gains are modest due to over-exploration; they take more steps and use more tools, but are often distracted by conflicting or partially relevant web content, leading to incorrect answers. Compared to humans, these agents consume more resources yet achieve lower accuracy, largely due to inefficient source selection and an overreliance on text modalities. These findings highlight the need for search agents capable of robust search and reasoning across diverse modalities in noisy web environments, making MERRIN a valuable testbed for evaluating such capabilities.</p> + https://arxiv.org/abs/2604.13418 + Wed, 15 Apr 2026 00:00:00 +0000 + + + OneHOI: Unifying Human-Object Interaction Generation and Editing + https://arxiv.org/abs/2604.14062 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14062.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiun Tian Hoe, Weipeng Hu, Xudong Jiang, Yap-Peng Tan, Chee Seng Chan</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Human-Object Interaction (HOI) modelling captures how humans act upon and relate to objects, typically expressed as <person, action, object> triplets. Existing approaches split into two disjoint families: HOI generation synthesises scenes from structured triplets and layout, but fails to integrate mixed conditions like HOI and object-only entities; and HOI editing modifies interactions via text, yet struggles to decouple pose from physical contact and scale to multiple interactions. We introduce OneHOI, a unified diffusion transformer framework that consolidates HOI generation and editing into a single conditional denoising process driven by shared structured interaction representations. At its core, the Relational Diffusion Transformer (R-DiT) models verb-mediated relations through role- and instance-aware HOI tokens, layout-based spatial Action Grounding, a Structured HOI Attention to enforce interaction topology, and HOI RoPE to disentangle multi-HOI scenes. Trained jointly with modality dropout on our HOI-Edit-44K, along with HOI and object-centric datasets, OneHOI supports layout-guided, layout-free, arbitrary-mask, and mixed-condition control, achieving state-of-the-art results across both HOI generation and editing. Code is available at https://jiuntian.github.io/OneHOI/.</p> + https://arxiv.org/abs/2604.14062 + Wed, 15 Apr 2026 00:00:00 +0000 + + + UI-Zoomer: Uncertainty-Driven Adaptive Zoom-In for GUI Grounding + https://arxiv.org/abs/2604.14113 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14113.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Fei Tang, Bofan Chen, Zhengxi Lu, Tongbo Chen, Songqin Nong, Tao Jiang, Wenhao Xu, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> GUI grounding, which localizes interface elements from screenshots given natural language queries, remains challenging for small icons and dense layouts. Test-time zoom-in methods improve localization by cropping and re-running inference at higher resolution, but apply cropping uniformly across all instances with fixed crop sizes, ignoring whether the model is actually uncertain on each case. We propose UI-Zoomer, a training-free adaptive zoom-in framework that treats both the trigger and scale of zoom-in as a prediction uncertainty quantification problem. A confidence-aware gate fuses spatial consensus among stochastic candidates with token-level generation confidence to selectively trigger zoom-in only when localization is uncertain. When triggered, an uncertainty-driven crop sizing module decomposes prediction variance into inter-sample positional spread and intra-sample box extent, deriving a per-instance crop radius via the law of total variance. Extensive experiments on ScreenSpot-Pro, UI-Vision, and ScreenSpot-v2 demonstrate consistent improvements over strong baselines across multiple model architectures, achieving gains of up to +13.4\%, +10.3\%, and +4.2\% respectively, with no additional training required.</p> + https://arxiv.org/abs/2604.14113 + Wed, 15 Apr 2026 00:00:00 +0000 + + + TIP: Token Importance in On-Policy Distillation + https://arxiv.org/abs/2604.14084 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14084.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuanda Xu, Hejian Sang, Zhengze Zhou, Ran He, Zhipeng Wang, Alborz Geramifard</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> On-policy knowledge distillation (OPD) trains a student on its own rollouts under token-level supervision from a teacher. Not all token positions matter equally, but existing views of token importance are incomplete. We ask a direct question: which tokens carry the most useful learning signal in OPD? Our answer is that informative tokens come from two regions: positions with high student entropy, and positions with low student entropy plus high teacher--student divergence, where the student is overconfident and wrong. Empirically, student entropy is a strong first-order proxy: retaining 50% of tokens with entropy-based sampling matches or exceeds all-token training while reducing peak memory by up to 47%. But entropy alone misses a second important region. When we isolate low-entropy, high-divergence tokens, training on fewer than 10% of all tokens nearly matches full-token baselines, showing that overconfident tokens carry dense corrective signal despite being nearly invisible to entropy-only rules. We organize these findings with TIP (Token Importance in on-Policy distillation), a two-axis taxonomy over student entropy and teacher--student divergence, and give a theoretical explanation for why entropy is useful yet structurally incomplete. This view motivates type-aware token selection rules that combine uncertainty and disagreement. We validate this picture across three teacher--student pairs spanning Qwen3, Llama, and Qwen2.5 on MATH-500 and AIME 2024/2025, and on the DeepPlanning benchmark for long-horizon agentic planning, where Q3-only training on <20% of tokens surpasses full-token OPD. Our experiments are implemented by extending the OPD repository https://github.com/HJSang/OPSD_OnPolicyDistillation, which supports memory-efficient distillation of larger models under limited GPU budgets.</p> + https://arxiv.org/abs/2604.14084 + Wed, 15 Apr 2026 00:00:00 +0000 + + + Feed-Forward 3D Scene Modeling: A Problem-Driven Perspective + https://arxiv.org/abs/2604.14025 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14025.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Weijie Wang, Qihang Cao, Sensen Gao, Donny Y. Chen, Haofei Xu, Wenjing Bian, Songyou Peng, Tat-Jen Cham, Chuanxia Zheng, Andreas Geiger, Jianfei Cai, Jia-Wang Bian, Bohan Zhuang</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Reconstructing 3D representations from 2D inputs is a fundamental task in computer vision and graphics, serving as a cornerstone for understanding and interacting with the physical world. While traditional methods achieve high fidelity, they are limited by slow per-scene optimization or category-specific training, which hinders their practical deployment and scalability. Hence, generalizable feed-forward 3D reconstruction has witnessed rapid development in recent years. By learning a model that maps images directly to 3D representations in a single forward pass, these methods enable efficient reconstruction and robust cross-scene generalization. Our survey is motivated by a critical observation: despite the diverse geometric output representations, ranging from implicit fields to explicit primitives, existing feed-forward approaches share similar high-level architectural patterns, such as image feature extraction backbones, multi-view information fusion mechanisms, and geometry-aware design principles. Consequently, we abstract away from these representation differences and instead focus on model design, proposing a novel taxonomy centered on model design strategies that are agnostic to the output format. Our proposed taxonomy organizes the research directions into five key problems that drive recent research development: feature enhancement, geometry awareness, model efficiency, augmentation strategies and temporal-aware models. To support this taxonomy with empirical grounding and standardized evaluation, we further comprehensively review related benchmarks and datasets, and extensively discuss and categorize real-world applications based on feed-forward 3D models. Finally, we outline future directions to address open challenges such as scalability, evaluation standards, and world modeling.</p> + https://arxiv.org/abs/2604.14025 + Wed, 15 Apr 2026 00:00:00 +0000 + + + TREX: Automating LLM Fine-tuning via Agent-Driven Tree-based Exploration + https://arxiv.org/abs/2604.14116 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14116.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zerun Ma, Guoqiang Wang, Xinchen Xie, Yicheng Chen, He Du, Bowen Li, Yanan Sun, Wenran Liu, Kai Chen, Yining Li</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> While Large Language Models (LLMs) have empowered AI research agents to perform isolated scientific tasks, automating complex, real-world workflows, such as LLM training, remains a significant challenge. In this paper, we introduce TREX, a multi-agent system that automates the entire LLM training life-cycle. By orchestrating collaboration between two core modules-the Researcher and the Executor-the system seamlessly performs requirement analysis, open-domain literature and data research, formulation of training strategies, preparation of data recipes, and model training and evaluation. The multi-round experimental process is modeled as a search tree, enabling the system to efficiently plan exploration paths, reuse historical results, and distill high-level insights from iterative trials. To evaluate the capability of automated LLM training, we construct FT-Bench, a benchmark comprising 10 tasks derived from real-world scenarios, ranging from optimizing fundamental model capabilities to enhancing performance on domain-specific tasks. Experimental results demonstrate that the TREX agent consistently optimizes model performance on target tasks.</p> + https://arxiv.org/abs/2604.14116 + Wed, 15 Apr 2026 00:00:00 +0000 + + + LangFlow: Continuous Diffusion Rivals Discrete in Language Modeling + https://arxiv.org/abs/2604.11748 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.11748.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuxin Chen, Chumeng Liang, Hangke Sui, Ruihan Guo, Chaoran Cheng, Jiaxuan You, Ge Liu</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> Continuous diffusion has been the foundation of high-fidelity, controllable, and few-step generation of many data modalities such as images. However, in language modeling, prior continuous diffusion language models (DLMs) lag behind discrete counterparts due to the sparse data space and the underexplored design space. In this work, we close this gap with LangFlow, the first continuous DLM to rival discrete diffusion, by connecting embedding-space DLMs to Flow Matching via Bregman divergence, alongside three key innovations: (1) we derive a novel ODE-based NLL bound for principled evaluation of continuous flow-based language models; (2) we propose an information-uniform principle for setting the noise schedule, which motivates a learnable noise scheduler based on a Gumbel distribution; and (3) we revise prior training protocols by incorporating self-conditioning, as we find it improves both likelihood and sample quality of embedding-space DLMs with effects substantially different from discrete diffusion. Putting everything together, LangFlow rivals top discrete DLMs on both the perplexity (PPL) and the generative perplexity (Gen. PPL), reaching a PPL of 30.0 on LM1B and 24.6 on OpenWebText. It even exceeds autoregressive baselines in zero-shot transfer on 4 out of 7 benchmarks. LangFlow provides the first clear evidence that continuous diffusion is a promising paradigm for language modeling. Homepage: https://github.com/nealchen2003/LangFlow</p> + https://arxiv.org/abs/2604.11748 + Wed, 15 Apr 2026 00:00:00 +0000 + + + Free Geometry: Refining 3D Reconstruction from Longer Versions of Itself + https://arxiv.org/abs/2604.14048 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14048.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuhang Dai, Xingyi Yang</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Feed-forward 3D reconstruction models are efficient but rigid: once trained, they perform inference in a zero-shot manner and cannot adapt to the test scene. As a result, visually plausible reconstructions often contain errors, particularly under occlusions, specularities, and ambiguous cues. To address this, we introduce Free Geometry, a framework that enables feed-forward 3D reconstruction models to self-evolve at test time without any 3D ground truth. Our key insight is that, when the model receives more views, it produces more reliable and view-consistent reconstructions. Leveraging this property, given a testing sequence, we mask a subset of frames to construct a self-supervised task. Free Geometry enforces cross-view feature consistency between representations from full and partial observations, while maintaining the pairwise relations implied by the held-out frames. This self-supervision allows for fast recalibration via lightweight LoRA updates, taking less than 2 minutes per dataset on a single GPU. Our approach consistently improves state-of-the-art foundation models, including Depth Anything 3 and VGGT, across 4 benchmark datasets, yielding an average improvement of 3.73% in camera pose accuracy and 2.88% in point map prediction. Code is available at https://github.com/hiteacherIamhumble/Free-Geometry .</p> + https://arxiv.org/abs/2604.14048 + Wed, 15 Apr 2026 00:00:00 +0000 + + + HiVLA: A Visual-Grounded-Centric Hierarchical Embodied Manipulation System + https://arxiv.org/abs/2604.14125 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14125.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tianshuo Yang, Guanyu Chen, Yutian Chen, Zhixuan Liang, Yitian Liu, Zanxin Chen, Chunpu Xu, Haotian Liang, Jiangmiao Pang, Yao Mu, Ping Luo</p><p><b>Upvotes:</b> 18</p><p><b>Summary:</b> While end-to-end Vision-Language-Action (VLA) models offer a promising paradigm for robotic manipulation, fine-tuning them on narrow control data often compromises the profound reasoning capabilities inherited from their base Vision-Language Models (VLMs). To resolve this fundamental trade-off, we propose HiVLA, a visual-grounded-centric hierarchical framework that explicitly decouples high-level semantic planning from low-level motor control. In high-level part, a VLM planner first performs task decomposition and visual grounding to generate structured plans, comprising a subtask instruction and a precise target bounding box. Then, to translate this plan into physical actions, we introduce a flow-matching Diffusion Transformer (DiT) action expert in low-level part equipped with a novel cascaded cross-attention mechanism. This design sequentially fuses global context, high-resolution object-centric crops and skill semantics, enabling the DiT to focus purely on robust execution. Our decoupled architecture preserves the VLM's zero-shot reasoning while allowing independent improvement of both components. Extensive experiments in simulation and the real world demonstrate that HiVLA significantly outperforms state-of-the-art end-to-end baselines, particularly excelling in long-horizon skill composition and the fine-grained manipulation of small objects in cluttered scenes.</p> + https://arxiv.org/abs/2604.14125 + Wed, 15 Apr 2026 00:00:00 +0000 + + + From P(y|x) to P(y): Investigating Reinforcement Learning in Pre-train Space + https://arxiv.org/abs/2604.14142 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14142.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuqiao Tan, Minzheng Wang, Bo Liu, Zichen Liu, Tian Liang, Shizhu He, Jun Zhao, Kang Liu</p><p><b>Upvotes:</b> 26</p><p><b>Summary:</b> While reinforcement learning with verifiable rewards (RLVR) significantly enhances LLM reasoning by optimizing the conditional distribution P(y|x), its potential is fundamentally bounded by the base model's existing output distribution. Optimizing the marginal distribution P(y) in the Pre-train Space addresses this bottleneck by encoding reasoning ability and preserving broad exploration capacity. Yet, conventional pre-training relies on static corpora for passive learning, leading to a distribution shift that hinders targeted reasoning enhancement. In this paper, we introduce PreRL (Pre-train Space RL), which applies reward-driven online updates directly to P(y). We theoretically and empirically validate the strong gradient alignment between log P(y) and log P(y|x), establishing PreRL as a viable surrogate for standard RL. Furthermore, we uncover a critical mechanism: Negative Sample Reinforcement (NSR) within PreRL serves as an exceptionally effective driver for reasoning. NSR-PreRL rapidly prunes incorrect reasoning spaces while stimulating endogenous reflective behaviors, increasing transition and reflection thoughts by 14.89x and 6.54x, respectively. Leveraging these insights, we propose Dual Space RL (DSRL), a Policy Reincarnation strategy that initializes models with NSR-PreRL to expand the reasoning horizon before transitioning to standard RL for fine-grained optimization. Extensive experiments demonstrate that DSRL consistently outperforms strong baselines, proving that pre-train space pruning effectively steers the policy toward a refined correct reasoning subspace.</p> + https://arxiv.org/abs/2604.14142 + Wed, 15 Apr 2026 00:00:00 +0000 + + + Memory Transfer Learning: How Memories are Transferred Across Domains in Coding Agents + https://arxiv.org/abs/2604.14004 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14004.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Kangsan Kim, Minki Kang, Taeil Kim, Yanlai Yang, Mengye Ren, Sung Ju Hwang</p><p><b>Upvotes:</b> 28</p><p><b>Summary:</b> Memory-based self-evolution has emerged as a promising paradigm for coding agents. However, existing approaches typically restrict memory utilization to homogeneous task domains, failing to leverage the shared infrastructural foundations, such as runtime environments and programming languages, that exist across diverse real-world coding problems. To address this limitation, we investigate Memory Transfer Learning (MTL) by harnessing a unified memory pool from heterogeneous domains. We evaluate performance across 6 coding benchmarks using four memory representations, ranging from concrete traces to abstract insights. Our experiments demonstrate that cross-domain memory improves average performance by 3.7\%, primarily by transferring meta-knowledge, such as validation routines, rather than task-specific code. Importantly, we find that abstraction dictates transferability; high-level insights generalize well, whereas low-level traces often induce negative transfer due to excessive specificity. Furthermore, we show that transfer effectiveness scales with the size of the memory pool, and memory can be transferred even between different models. Our work establishes empirical design principles for expanding memory utilization beyond single-domain silos. Project page: https://memorytransfer.github.io/</p> + https://arxiv.org/abs/2604.14004 + Wed, 15 Apr 2026 00:00:00 +0000 + + + SpatialEvo: Self-Evolving Spatial Intelligence via Deterministic Geometric Environments + https://arxiv.org/abs/2604.14144 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14144.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dinging Li, Yingxiu Zhao, Xinrui Cheng, Kangheng Lin, Hongbo Peng, Hongxing Li, Zixuan Wang, Yuhong Dai, Haodong Li, Jia Wang, Yukang Shi, Liang Zhao, Jianjian Sun, Zheng Ge, Xiangyu Zhang, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen</p><p><b>Upvotes:</b> 62</p><p><b>Summary:</b> Spatial reasoning over three-dimensional scenes is a core capability for embodied intelligence, yet continuous model improvement remains bottlenecked by the cost of geometric annotation. The self-evolving paradigm offers a promising path, but its reliance on model consensus to construct pseudo-labels causes training to reinforce rather than correct the model's own geometric errors. We identify a property unique to 3D spatial reasoning that circumvents this limitation: ground truth is a deterministic consequence of the underlying geometry, computable exactly from point clouds and camera poses without any model involvement. Building on this insight, we present SpatialEvo, a self-evolving framework for 3D spatial reasoning, centered on the Deterministic Geometric Environment (DGE). The DGE formalizes 16 spatial reasoning task categories under explicit geometric validation rules and converts unannotated 3D scenes into zero-noise interactive oracles, replacing model consensus with objective physical feedback. A single shared-parameter policy co-evolves across questioner and solver roles under DGE constraints: the questioner generates physically valid spatial questions grounded in scene observations, while the solver derives precise answers against DGE-verified ground truth. A task-adaptive scheduler endogenously concentrates training on the model's weakest categories, producing a dynamic curriculum without manual design. Experiments across nine benchmarks demonstrate that SpatialEvo achieves the highest average score at both 3B and 7B scales, with consistent gains on spatial reasoning benchmarks and no degradation on general visual understanding.</p> + https://arxiv.org/abs/2604.14144 + Wed, 15 Apr 2026 00:00:00 +0000 + + + HY-World 2.0: A Multi-Modal World Model for Reconstructing, Generating, and Simulating 3D Worlds + https://arxiv.org/abs/2604.14268 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14268.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Team HY-World, Chenjie Cao, Xuhui Zuo, Zhenwei Wang, Yisu Zhang, Junta Wu, Zhenyang Liu, Yuning Gong, Yang Liu, Bo Yuan, Chao Zhang, Coopers Li, Dongyuan Guo, Fan Yang, Haiyu Zhang, Hang Cao, Jianchen Zhu, Jiaxin Lin, Jie Xiao, Jihong Zhang, Junlin Yu, Lei Wang, Lifu Wang, Lilin Wang, Linus, Minghui Chen, Peng He, Penghao Zhao, Qi Chen, Rui Chen, Rui Shao, Sicong Liu, Wangchen Qin, Xiaochuan Niu, Xiang Yuan, Yi Sun, Yifei Tang, Yifu Sun, Yihang Lian, Yonghao Tan, Yuhong Liu, Yuyang Yin, Zhiyuan Min, Tengfei Wang, Chunchao Guo</p><p><b>Upvotes:</b> 90</p><p><b>Summary:</b> We introduce HY-World 2.0, a multi-modal world model framework that advances our prior project HY-World 1.0. HY-World 2.0 accommodates diverse input modalities, including text prompts, single-view images, multi-view images, and videos, and produces 3D world representations. With text or single-view image inputs, the model performs world generation, synthesizing high-fidelity, navigable 3D Gaussian Splatting (3DGS) scenes. This is achieved through a four-stage method: a) Panorama Generation with HY-Pano 2.0, b) Trajectory Planning with WorldNav, c) World Expansion with WorldStereo 2.0, and d) World Composition with WorldMirror 2.0. Specifically, we introduce key innovations to enhance panorama fidelity, enable 3D scene understanding and planning, and upgrade WorldStereo, our keyframe-based view generation model with consistent memory. We also upgrade WorldMirror, a feed-forward model for universal 3D prediction, by refining model architecture and learning strategy, enabling world reconstruction from multi-view images or videos. Also, we introduce WorldLens, a high-performance 3DGS rendering platform featuring a flexible engine-agnostic architecture, automatic IBL lighting, efficient collision detection, and training-rendering co-design, enabling interactive exploration of 3D worlds with character support. Extensive experiments demonstrate that HY-World 2.0 achieves state-of-the-art performance on several benchmarks among open-source approaches, delivering results comparable to the closed-source model Marble. We release all model weights, code, and technical details to facilitate reproducibility and support further research on 3D world models.</p> + https://arxiv.org/abs/2604.14268 + Wed, 15 Apr 2026 00:00:00 +0000 + + + Seedance 2.0: Advancing Video Generation for World Complexity + https://arxiv.org/abs/2604.14148 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14148.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Team Seedance, De Chen, Liyang Chen, Xin Chen, Ying Chen, Zhuo Chen, Zhuowei Chen, Feng Cheng, Tianheng Cheng, Yufeng Cheng, Mojie Chi, Xuyan Chi, Jian Cong, Qinpeng Cui, Fei Ding, Qide Dong, Yujiao Du, Haojie Duanmu, Junliang Fan, Jiarui Fang, Jing Fang, Zetao Fang, Chengjian Feng, Yu Gao, Diandian Gu, Dong Guo, Hanzhong Guo, Qiushan Guo, Boyang Hao, Hongxiang Hao, Haoxun He, Jiaao He, Qian He, Tuyen Hoang, Heng Hu, Ruoqing Hu, Yuxiang Hu, Jiancheng Huang, Weilin Huang, Zhaoyang Huang, Zhongyi Huang, Jishuo Jin, Ming Jing, Ashley Kim, Shanshan Lao, Yichong Leng, Bingchuan Li, Gen Li, Haifeng Li, Huixia Li, Jiashi Li, Ming Li, Xiaojie Li, Xingxing Li, Yameng Li, Yiying Li, Yu Li, Yueyan Li, Chao Liang, Han Liang, Jianzhong Liang, Ying Liang, Wang Liao, J. H. Lien, Shanchuan Lin, Xi Lin, Feng Ling, Yue Ling, Fangfang Liu, Jiawei Liu, Jihao Liu, Jingtuo Liu, Shu Liu, Sichao Liu, Wei Liu, Xue Liu, Zuxi Liu, Ruijie Lu, Lecheng Lyu, Jingting Ma, Tianxiang Ma, Xiaonan Nie, Jingzhe Ning, Junjie Pan, Xitong Pan, Ronggui Peng, Xueqiong Qu, Yuxi Ren, Yuchen Shen, Guang Shi, Lei Shi, Yinglong Song, Fan Sun, Li Sun, Renfei Sun, Wenjing Tang, Boyang Tao, Zirui Tao, Dongliang Wang, Feng Wang, Hulin Wang, Ke Wang, Qingyi Wang, Rui Wang, Shuai Wang, Shulei Wang, Weichen Wang, Xuanda Wang, Yanhui Wang, Yue Wang, Yuping Wang, Yuxuan Wang, Zijie Wang, Ziyu Wang, Guoqiang Wei, Meng Wei, Di Wu, Guohong Wu, Hanjie Wu, Huachao Wu, Jian Wu, Jie Wu, Ruolan Wu, Shaojin Wu, Xiaohu Wu, Xinglong Wu, Yonghui Wu, Ruiqi Xia, Xin Xia, Xuefeng Xiao, Shuang Xu, Bangbang Yang, Jiaqi Yang, Runkai Yang, Tao Yang, Yihang Yang, Zhixian Yang, Ziyan Yang, Fulong Ye, Bingqian Yi, Xing Yin, Yongbin You, Linxiao Yuan, Weihong Zeng, Xuejiao Zeng, Yan Zeng, Siyu Zhai, Zhonghua Zhai, Bowen Zhang, Chenlin Zhang, Heng Zhang, Jun Zhang, Manlin Zhang, Peiyuan Zhang, Shuo Zhang, Xiaohe Zhang, Xiaoying Zhang, Xinyan Zhang, Xinyi Zhang, Yichi Zhang, Zixiang Zhang, Haiyu Zhao, Huating Zhao, Liming Zhao, Yian Zhao, Guangcong Zheng, Jianbin Zheng, Xiaozheng Zheng, Zerong Zheng, Kuan Zhu, Feilong Zuo</p><p><b>Upvotes:</b> 140</p><p><b>Summary:</b> Seedance 2.0 is a new native multi-modal audio-video generation model, officially released in China in early February 2026. Compared with its predecessors, Seedance 1.0 and 1.5 Pro, Seedance 2.0 adopts a unified, highly efficient, and large-scale architecture for multi-modal audio-video joint generation. This allows it to support four input modalities: text, image, audio, and video, by integrating one of the most comprehensive suites of multi-modal content reference and editing capabilities available in the industry to date. It delivers substantial, well-rounded improvements across all key sub-dimensions of video and audio generation. In both expert evaluations and public user tests, the model has demonstrated performance on par with the leading levels in the field. Seedance 2.0 supports direct generation of audio-video content with durations ranging from 4 to 15 seconds, with native output resolutions of 480p and 720p. For multi-modal inputs as reference, its current open platform supports up to 3 video clips, 9 images, and 3 audio clips. In addition, we provide Seedance 2.0 Fast version, an accelerated variant of Seedance 2.0 designed to boost generation speed for low-latency scenarios. Seedance 2.0 has delivered significant improvements to its foundational generation capabilities and multi-modal generation performance, bringing an enhanced creative experience for end users.</p> + https://arxiv.org/abs/2604.14148 + Wed, 15 Apr 2026 00:00:00 +0000 + + + LongAct: Harnessing Intrinsic Activation Patterns for Long-Context Reinforcement Learning + https://arxiv.org/abs/2604.14922 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14922.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bowen Ping, Zijun Chen, Tingfeng Hui, Qize Yu, Chenxuan Li, Junchi Yan, Baobao Chang</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Reinforcement Learning (RL) has emerged as a critical driver for enhancing the reasoning capabilities of Large Language Models (LLMs). While recent advancements have focused on reward engineering or data synthesis, few studies exploit the model's intrinsic representation characteristics to guide the training process. In this paper, we first observe the presence of high-magnitude activations within the query and key vectors when processing long contexts. Drawing inspiration from model quantization -- which establishes the criticality of such high-magnitude activations -- and the insight that long-context reasoning inherently exhibits a sparse structure, we hypothesize that these weights serve as the pivotal drivers for effective model optimization. Based on this insight, we propose LongAct, a strategy that shifts from uniform to saliency-guided sparse updates. By selectively updating only the weights associated with these significant activations, LongAct achieves an approximate 8% improvement on LongBench v2 and enhances generalization on the RULER benchmark. Furthermore, our method exhibits remarkable universality, consistently boosting performance across diverse RL algorithms such as GRPO and DAPO. Extensive ablation studies suggest that focusing on these salient features is key to unlocking long-context potential.</p> + https://arxiv.org/abs/2604.14922 + Thu, 16 Apr 2026 00:00:00 +0000 + + + Switch-KD: Visual-Switch Knowledge Distillation for Vision-Language Models + https://arxiv.org/abs/2604.14629 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14629.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoyi Sun, Xiaoxiao Wang, Ning Mao, Qian Wang, Lifu Mu, Wen Zheng, Tao Wei, Wei Chen</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Vision-Language Models (VLMs) have shown remarkable capabilities in joint vision-language understanding, but their large scale poses significant challenges for deployment in resource-constrained scenarios. Knowledge Distillation (KD) offers a viable way to improve model capabilities without increasing model size or data requirements, making deployment more efficient. However, applying KD to VLMs is challenged by modality-specific supervision: although multimodal knowledge in VLMs is fused within the language space, current methods supervise each modality separately without explicitly addressing multimodal alignment, leading to inconsistent multimodal knowledge transfer. To address this, we propose Switch-KD, a visual-switch distillation framework that unifies vision-language knowledge transfer within a shared text-probability space. Switch-KD comprises two key components: (1) Visual-Switch Distillation, which switches the student's visual outputs into the teacher's language pathway to construct cross-modal probabilistic references for implicit visual knowledge transfer; and (2) Dynamic Bi-directional Logits Difference (DBiLD) loss, which adaptively aligns informative probability regions while preserving the distributional structures of teacher and student through bidirectional supervision. Guided by Switch-KD, a 0.5B TinyLLaVA effectively distills rich multimodal knowledge from its 3B teacher, yielding an average improvement of 3.6 points across 10 multimodal benchmarks without any architectural modification.</p> + https://arxiv.org/abs/2604.14629 + Thu, 16 Apr 2026 00:00:00 +0000 + + + LeapAlign: Post-Training Flow Matching Models at Any Generation Step by Building Two-Step Trajectories + https://arxiv.org/abs/2604.15311 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.15311.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhanhao Liang, Tao Yang, Jie Wu, Chengjian Feng, Liang Zheng</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> This paper focuses on the alignment of flow matching models with human preferences. A promising way is fine-tuning by directly backpropagating reward gradients through the differentiable generation process of flow matching. However, backpropagating through long trajectories results in prohibitive memory costs and gradient explosion. Therefore, direct-gradient methods struggle to update early generation steps, which are crucial for determining the global structure of the final image. To address this issue, we introduce LeapAlign, a fine-tuning method that reduces computational cost and enables direct gradient propagation from reward to early generation steps. Specifically, we shorten the long trajectory into only two steps by designing two consecutive leaps, each skipping multiple ODE sampling steps and predicting future latents in a single step. By randomizing the start and end timesteps of the leaps, LeapAlign leads to efficient and stable model updates at any generation step. To better use such shortened trajectories, we assign higher training weights to those that are more consistent with the long generation path. To further enhance gradient stability, we reduce the weights of gradient terms with large magnitude, instead of completely removing them as done in previous works. When fine-tuning the Flux model, LeapAlign consistently outperforms state-of-the-art GRPO-based and direct-gradient methods across various metrics, achieving superior image quality and image-text alignment.</p> + https://arxiv.org/abs/2604.15311 + Thu, 16 Apr 2026 00:00:00 +0000 + + + UniDoc-RL: Coarse-to-Fine Visual RAG with Hierarchical Actions and Dense Rewards + https://arxiv.org/abs/2604.14967 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14967.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jun Wang, Shuo Tan, Zelong Sun, Tiancheng Gu, Yongle Zhao, Ziyong Feng, Kaicheng Yang, Cewu Lu</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Retrieval-Augmented Generation (RAG) extends Large Vision-Language Models (LVLMs) with external visual knowledge. However, existing visual RAG systems typically rely on generic retrieval signals that overlook the fine-grained visual semantics essential for complex reasoning. To address this limitation, we propose UniDoc-RL, a unified reinforcement learning framework in which an LVLM agent jointly performs retrieval, reranking, active visual perception, and reasoning. UniDoc-RL formulates visual information acquisition as a sequential decision-making problem with a hierarchical action space. Specifically, it progressively refines visual evidence from coarse-grained document retrieval to fine-grained image selection and active region cropping, allowing the model to suppress irrelevant content and attend to information-dense regions. For effective end-to-end training, we introduce a dense multi-reward scheme that provides task-aware supervision for each action. Based on Group Relative Policy Optimization (GRPO), UniDoc-RL aligns agent behavior with multiple objectives without relying on a separate value network. To support this training paradigm, we curate a comprehensive dataset of high-quality reasoning trajectories with fine-grained action annotations. Experiments on three benchmarks demonstrate that UniDoc-RL consistently surpasses state-of-the-art baselines, yielding up to 17.7% gains over prior RL-based methods.</p> + https://arxiv.org/abs/2604.14967 + Thu, 16 Apr 2026 00:00:00 +0000 + + + GlobalSplat: Efficient Feed-Forward 3D Gaussian Splatting via Global Scene Tokens + https://arxiv.org/abs/2604.15284 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.15284.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Roni Itkin, Noam Issachar, Yehonatan Keypur, Yehonatan Keypur, Anpei Chen, Sagie Benaim</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> The efficient spatial allocation of primitives serves as the foundation of 3D Gaussian Splatting, as it directly dictates the synergy between representation compactness, reconstruction speed, and rendering fidelity. Previous solutions, whether based on iterative optimization or feed-forward inference, suffer from significant trade-offs between these goals, mainly due to the reliance on local, heuristic-driven allocation strategies that lack global scene awareness. Specifically, current feed-forward methods are largely pixel-aligned or voxel-aligned. By unprojecting pixels into dense, view-aligned primitives, they bake redundancy into the 3D asset. As more input views are added, the representation size increases and global consistency becomes fragile. To this end, we introduce GlobalSplat, a framework built on the principle of align first, decode later. Our approach learns a compact, global, latent scene representation that encodes multi-view input and resolves cross-view correspondences before decoding any explicit 3D geometry. Crucially, this formulation enables compact, globally consistent reconstructions without relying on pretrained pixel-prediction backbones or reusing latent features from dense baselines. Utilizing a coarse-to-fine training curriculum that gradually increases decoded capacity, GlobalSplat natively prevents representation bloat. On RealEstate10K and ACID, our model achieves competitive novel-view synthesis performance while utilizing as few as 16K Gaussians, significantly less than required by dense pipelines, obtaining a light 4MB footprint. Further, GlobalSplat enables significantly faster inference than the baselines, operating under 78 milliseconds in a single forward pass. Project page is available at https://r-itk.github.io/globalsplat/</p> + https://arxiv.org/abs/2604.15284 + Thu, 16 Apr 2026 00:00:00 +0000 + + + RAD-2: Scaling Reinforcement Learning in a Generator-Discriminator Framework + https://arxiv.org/abs/2604.15308 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.15308.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hao Gao, Shaoyu Chen, Yifan Zhu, Yuehao Song, Wenyu Liu, Qian Zhang, Xinggang Wang</p><p><b>Upvotes:</b> 26</p><p><b>Summary:</b> High-level autonomous driving requires motion planners capable of modeling multimodal future uncertainties while remaining robust in closed-loop interactions. Although diffusion-based planners are effective at modeling complex trajectory distributions, they often suffer from stochastic instabilities and the lack of corrective negative feedback when trained purely with imitation learning. To address these issues, we propose RAD-2, a unified generator-discriminator framework for closed-loop planning. Specifically, a diffusion-based generator is used to produce diverse trajectory candidates, while an RL-optimized discriminator reranks these candidates according to their long-term driving quality. This decoupled design avoids directly applying sparse scalar rewards to the full high-dimensional trajectory space, thereby improving optimization stability. To further enhance reinforcement learning, we introduce Temporally Consistent Group Relative Policy Optimization, which exploits temporal coherence to alleviate the credit assignment problem. In addition, we propose On-policy Generator Optimization, which converts closed-loop feedback into structured longitudinal optimization signals and progressively shifts the generator toward high-reward trajectory manifolds. To support efficient large-scale training, we introduce BEV-Warp, a high-throughput simulation environment that performs closed-loop evaluation directly in Bird's-Eye View feature space via spatial warping. RAD-2 reduces the collision rate by 56% compared with strong diffusion-based planners. Real-world deployment further demonstrates improved perceived safety and driving smoothness in complex urban traffic.</p> + https://arxiv.org/abs/2604.15308 + Thu, 16 Apr 2026 00:00:00 +0000 + + + DR^{3}-Eval: Towards Realistic and Reproducible Deep Research Evaluation + https://arxiv.org/abs/2604.14683 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2604.14683.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qianqian Xie, Qingheng Xiong, He Zhu, Tiantian Xia, Xueming Han, Fanyu Meng, Jiakai Wang, Zhiqi Bai, Chengkang Jiang, Zhaohui Wang, Yubin Guo, Yuqing Wen, Jiayang Mao, Zijie Zhang, Shihao Li, Yanghai Wang, Yuxiang Ren, Junlan Feng, Jiaheng Liu</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> Deep Research Agents (DRAs) aim to solve complex, long-horizon research tasks involving planning, retrieval, multimodal understanding, and report generation, yet their evaluation remains challenging due to dynamic web environments and ambiguous task definitions. We propose DR^{3}-Eval, a realistic and reproducible benchmark for evaluating deep research agents on multimodal, multi-file report generation. DR^{3}-Eval is constructed from authentic user-provided materials and paired with a per-task static research sandbox corpus that simulates open-web complexity while remaining fully verifiable, containing supportive documents, distractors, and noise. Moreover, we introduce a multi-dimensional evaluation framework measuring Information Recall, Factual Accuracy, Citation Coverage, Instruction Following, and Depth Quality, and validate its alignment with human judgments. Experiments with our developed multi-agent system DR^{3}-Agent based on multiple state-of-the-art language models demonstrate that DR^{3}-Eval is highly challenging and reveals critical failure modes in retrieval robustness and hallucination control. Our code and data are publicly available.</p> + https://arxiv.org/abs/2604.14683 + Thu, 16 Apr 2026 00:00:00 +0000