From 225b80df1ae72504edce7e4a1c10a086627c4d67 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" Date: Thu, 25 Jun 2026 00:23:54 +0000 Subject: [PATCH] bot: update RSS feed --- feed.xml | 448 +++++++++++++++++-------------------------------------- 1 file changed, 133 insertions(+), 315 deletions(-) diff --git a/feed.xml b/feed.xml index 3ca79ae..31df0b2 100644 --- a/feed.xml +++ b/feed.xml @@ -7,391 +7,209 @@ http://www.rssboard.org/rss-specification python-feedgen en - Wed, 24 Jun 2026 00:18:54 +0000 + Thu, 25 Jun 2026 00:23:54 +0000 - SkillHarness: Harnessing Safe Skills for Computer-Use Agents - https://arxiv.org/abs/2606.20636 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20636.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yurun Chen, Biao Yi, Keting Yin, Shengyu Zhang</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> Computer-Use Agents (CUAs) are increasingly deployed in dynamic interactive environments, creating a growing need for continual skill learning during interaction. Recent approaches address this challenge by learning reusable skills from successful trajectories. However, these skill learning methods largely assume static and safe environments, overlooking risks from adversarial interactions (e.g., prompt injections) and environmental dynamics (e.g., pop-ups). In dynamic settings, such assumptions can lead to risky skill learning and brittle execution, undermining the reliability of CUAs. This raises the question: how can CUAs learn and use skills safely in dynamic environments? To address this problem, we propose SkillHarness, a framework for safe skill harnessing in dynamic environments. SkillHarness moves beyond static skill abstractions by modeling skill learning and utilization as a safety-constrained interaction process. Specifically, we introduce the skill boundary that leverages multi-source supervision signals to identify safe skills from interaction trajectories, and construct self-improving safety constraints throughout the skill lifecycle. In addition, SkillHarness introduces selective skill reuse, where tasks are guided to decompose according to context and completed through the selective activation of skill subsets. Our experiments demonstrate that SkillHarness significantly reduces the unsafe rate of learned skills by 57.1% and consistently improves execution stability under dynamic environmental changes, outperforming existing baselines.</p> - https://arxiv.org/abs/2606.20636 - Tue, 02 Jun 2026 00:00:00 +0000 - - - Exploring the Design Space of Reward Backpropagation for Flow Matching - https://arxiv.org/abs/2606.11075 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.11075.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ruoyu Wang, Boye Niu, Xiangxin Zhou, Yushi Huang, Tongliang Liu, Chi Zhang</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Aligning text-to-image flow matching models with human preferences via direct reward backpropagation is sample-efficient but hampered by two well-known pathologies: activations cannot be stored across the full sampling trajectory at modern model scale, and chained Jacobian products across steps inflate the reward gradient as it travels back to early indices. Connector-based methods, such as LeapAlign, address these issues by replacing the full backward trajectory with a short pinned path, highlighting a useful decoupling between sampling and optimization. However, the quality of the resulting gradient depends on how accurately this short path approximates the full rollout, especially over long intervals. We propose FlowBP, a unified surrogate-trajectory framework that treats the backward trajectory itself as the design object. FlowBP keeps a no-gradient cached rollout for sampling, then builds a lightweight backward surrogate from cached and selectively re-forwarded velocities. This view separates four choices: the reward-model input, active set, integration weights, and bridge coupling, and recovers prior direct-gradient methods as particular settings. Within this framework, we instantiate three variants: FlowBP-Sparse uses sparse Euler reconstruction, FlowBP-Bridge adds controlled bridge coupling, and FlowBP-Lagrange raises the order of leap quadrature. All three bound memory by the active-set size and limit gradient chaining to at most one Jacobian factor. Across SD3.5-M, FLUX.1-dev, and FLUX.2-Klein-base on preference, quality, and compositional metrics, the three variants improve over direct-gradient baselines on most metrics.</p> - https://arxiv.org/abs/2606.11075 - Tue, 09 Jun 2026 16:36:54 +0000 - - - Notes2Skills: From Lab Notebooks to Certainty-Aware Scientific Agent Skills - https://arxiv.org/abs/2606.11897 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.11897.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shi Liu, Jiayao Chen, Chengwei Qin, Yanqing Hu, Jufan Zhang, Linyi Yang</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Scientific discovery workflows usually contain and rely heavily on lab notes, where researchers record observations, interpret uncertain results, and plan follow-up experiments. Such informative lab notes preserve evolving scientific reasoning and author uncertainty, rather than polished final results exhibited in publications, providing a valuable opportunity for AI to engage in scientific exploration at a more comprehensive and deeper level. However, most prior work on scientific text focuses on papers, protocols, or structured databases, leaving informal laboratory notes underexplored as inputs to AI agents for science. This gap matters because lab notes often intermingle validated observations, tentative judgments, and possible experimental next steps within the same passage. If these signals are conflated, an AI agent may mistake uncertain scientific judgments for confirmed conclusions or executable actions. To this end, we present Notes2Skills, a two-stage framework for turning lab notebooks into verifiable skills for scientific AI agents while preserving the author's certainty. Across seven conditions and three wet-lab sessions, Notes2Skills is the only configuration that neither mistakes uncertain notes for firm instructions nor discards firm ones. We show that certainty preservation is the missing piece between lab notebooks and reliable agent skills, opening a path toward safer AI co-scientist systems.</p> - https://arxiv.org/abs/2606.11897 + An Efficient Method for the Optimal Control of Microgrids Under Uncertainties using Local Reduction + https://arxiv.org/abs/2606.12345 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.12345.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Edoardo Scaccia, Eric C. Kerrigan, Anna Sadowska</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> The problem of optimal sizing and power scheduling in microgrids subject to uncertainties is well known to the control community. Commonly, the optimal control problem is cast as a mixed-integer program to model the logical constraints arising in energy storage systems, and is then solved approximately using numerical methods such as the scenario approach. In this paper, we propose and compare two formulations of a robust microgrid sizing and power scheduling optimal control problem with logical constraints and uncertainties in the user's power demand, solar power generation, grid electricity prices and battery efficiencies. The first formulation uses binary variables and big-M constraints, leading to a mixed-integer linear program. The second formulation casts the problem as a continuous nonlinear program through an exact smooth reformulation of the logical constraints, consisting of additional modelling variables and non-convex constraints. We then propose a novel local reduction algorithm, extending an existing method, to solve both problems. The two formulations are compared by evaluating the solutions returned by local reduction using 100,000-sample Monte Carlo simulations and achieve promising results, with both averaging feasibility rates above 90%.</p> + https://arxiv.org/abs/2606.12345 Wed, 10 Jun 2026 00:00:00 +0000 - DailyReport: An Open-ended Benchmark for Evaluating Search Agents on Daily Search Tasks - https://arxiv.org/abs/2606.12871 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.12871.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jingxuan Han, Wei Liu, Mingyang Zhu, Youpeng Wang, Ziwen Wang, Lin Qiu, Xuezhi Cao, Xunliang Cai, Zheren Fu, Licheng Zhang, Zhendong Mao</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Search Agents (SAs) typically leverage large language models (LLMs) to support complex information-seeking tasks by autonomously exploring web sources and synthesizing information into comprehensive responses. For SAs evaluation, prior benchmarks mainly focus on specialized tasks that are unlikely to arise in real-world user scenarios. Moreover, their reliance on coarse task-level rubrics often limits evaluation interpretability. To bridge this gap, we introduce DailyReport, an open-ended benchmark to evaluate SA capabilities on daily search tasks. It contains 150 open-ended tasks with 3,546 associated rubrics, capturing widely discussed and timely information demands of real-world users. Each task is decomposed into subtasks and evaluated with cascade rubrics across disentangled dimensions. Through cascade performance attribution and user-centric aggregation, we derive highly interpretable scores for each dimension, along with a user preference score. Our results on 17 agentic systems show that current systems still fall short of users' expectations. To facilitate future research, our dataset and code are made publicly available at https://github.com/AGI-Eval-Official/DailyReport.</p> - https://arxiv.org/abs/2606.12871 + LingxiDiagBench: A Multi-Agent Framework for Benchmarking LLMs in Chinese Psychiatric Consultation and Diagnosis + https://arxiv.org/abs/2602.09379 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2602.09379.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shihao Xu, Tiancheng Zhou, Jiatong Ma, Yanli Ding, Yiming Yan, Ming Xiao, Guoyi Li, Haiyang Geng, Yunyun Han, Jianhua Chen, Yafeng Deng</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Mental disorders are highly prevalent worldwide, but the shortage of psychiatrists and the inherent subjectivity of interview-based diagnosis create substantial barriers to timely and consistent mental-health assessment. Progress in AI-assisted psychiatric diagnosis is constrained by the absence of benchmarks that simultaneously provide realistic patient simulation, clinician-verified diagnostic labels, and support for dynamic multi-turn consultation. We present LingxiDiagBench, a large-scale multi-agent benchmark that evaluates LLMs on both static diagnostic inference and dynamic multi-turn psychiatric consultation in Chinese. At its core is LingxiDiag-16K, a dataset of 16,000 EMR-aligned synthetic consultation dialogues designed to reproduce real clinical demographic and diagnostic distributions across 12 ICD-10 psychiatric categories. Through extensive experiments across state-of-the-art LLMs, we establish key findings: (1) although LLMs achieve high accuracy on binary depression--anxiety classification (up to 92.3%), performance deteriorates substantially for depression--anxiety comorbidity recognition (43.0%) and 12-way differential diagnosis (28.5%); (2) dynamic consultation often underperforms static evaluation, indicating that ineffective information-gathering strategies significantly impair downstream diagnostic reasoning; (3) consultation quality assessed by LLM-as-a-Judge shows only moderate correlation with diagnostic accuracy, suggesting that well-structured questioning alone does not ensure correct diagnostic decisions. We release LingxiDiag-16K and the full evaluation framework to support reproducible research at https://github.com/Lingxi-mental-health/LingxiDiagBench.</p> + https://arxiv.org/abs/2602.09379 Thu, 11 Jun 2026 00:00:00 +0000 - FastMix: Fast Data Mixture Optimization via Gradient Descent - https://arxiv.org/abs/2606.14971 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.14971.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoru Tan, Sitong Wu, Yanfeng Chen, Jun Xia, Ruobing Xie, Bin Xia, Xingwu Sun, Xiaojuan Qi</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> While large and diverse datasets have driven recent advances in large models, identifying the optimal data mixture for pre-training and post-training remains a significant open problem. We address this challenge with FASTMIX, a novel framework that automates data mixture discovery while training only a single proxy model. Instead of relying on predefined heuristics or resource-intensive simulations, FASTMIX jointly optimizes mixture coefficients and model parameters, substantially improving efficiency and scalability over prior approaches. At the core of FASTMIX is a reformulation of mixture selection as a bilevel optimization problem. Under this reformulation, we show that optimizing mixture ratios is mathematically equivalent to assigning per-source loss weights under uniform source sampling. This embeds the mixture coefficients directly into the differentiable iterative optimization objective, enabling efficient, gradient-based optimization of both mixture and model. To solve the optimization problem, FASTMIX implements an approximate iterative optimization procedure, alternating between (i) updating model parameters on data sampled according to current mixture ratios (inner loop) and (ii) updating mixture ratios based on validation feedback (outer loop). Across pre- and post-training, FASTMIX outperforms baselines while drastically reducing search cost. Code (https://github.com/hrtan/fastmix)</p> - https://arxiv.org/abs/2606.14971 - Fri, 12 Jun 2026 00:00:00 +0000 - - - AC-ODM: Actor--Critic Online Data Mixing for Sample-Efficient LLM Pretraining - https://arxiv.org/abs/2505.23878 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2505.23878.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jing Ma, Chenhao Dang, Mingjie Liao</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Optimizing pretraining data composition is pivotal for LLM generalization. While dynamic mixing outperforms static strategies by capturing evolving training dynamics, current methods fail to reconcile computational efficiency with sample efficiency and structural flexibility for diverse pipelines.We introduce Actor--Critic Online Data Mixing (AC-ODM), which approaches data mixing from a reinforcement learning perspective with a parameterized policy that we theoretically prove to act as a dynamic linear surrogate maximizing the constructive interference of gradients. To enhance practical flexibility, AC-ODM supports two operational modes: (i) a proxy mode for fixed, pre-prepared corpora, where a policy learned on a small model is transferred to a larger target; and (ii) a non-proxy mode for direct end-to-end training from scratch without priors. Empirically, AC-ODM significantly outperforms prior methods in convergence speed and downstream accuracy across various architectures. On Pythia-1B, it reaches optimal validation perplexity using up to 66% fewer training steps than competitive baselines, delivering a 27.5% relative improvement in MMLU accuracy and a 2.23 x higher pass@1 on HumanEval, all while incurring a virtually negligible (0.4%) per-step wall-clock increase and only 2% additional memory overhead. Code is available at https://github.com/DANG-ai/AC-ODM.</p> - https://arxiv.org/abs/2505.23878 - Sun, 14 Jun 2026 00:00:00 +0000 - - - Comparing Linear Probes with Mahalanobis Cosine Similarity - https://arxiv.org/abs/2606.19603 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.19603.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhuofan Josh Ying, Peter Hase, Nikolaus Kriegeskorte</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Linear probes are widely used in interpretability research and often compared by cosine similarity. The Mahalanobis cosine similarity (MCS) between two directions, which reweights the inner product by test data covariance, is a natural task-aware refinement. Ying et al. (2026) report that a probe's MCS to a reference probe trained on the out-of-distribution (OOD) data near-perfectly linearly predicts the probe's OOD AUROC (R^2 = 0.98). Here, we extend this empirical finding across models, layers, and concept domains, and prove this general phenomenon in closed form: For balanced classes whose projections are Gaussian, OOD AUROC and MCS to the reference probe are linear because both are sigmoid-shaped functions of the probe's signal-to-noise ratio (SNR) on the test data. The theory also predicts when this linearity fails, which we verify empirically. MCS offers a theoretically grounded and empirically effective alternative to Euclidean cosine similarity for comparing linear probes.</p> - https://arxiv.org/abs/2606.19603 - Wed, 17 Jun 2026 00:00:00 +0000 - - - Deep Research in Physical Sciences: A Multi-Agent Framework and Comprehensive Benchmark - https://arxiv.org/abs/2606.18648 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.18648.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yigeng Jiang, Tengchao Yang, Taoyong Cui, Jiaxing Wan, Yuan Wang, Weida Wang, Zhiyu Liu, Chuyi Peng, Binzhao Luo, Maoli Gao, Huaihai Huang, Yuqianer Zeng, Ziyang Zheng, Dongchen Huang, Chao Chen, Zichao Liu, Weiping Shen, Shuchen Pu, Siyu Zhou, Runmin Ma, Yusong Hu, Fei Chao, Bo Zhang, Xiawu Zheng, Zifu Wang, Lei Bai, Yunqi Cai, Shufei Zhang</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Deep research agents are Large Language Model (LLM)-based systems designed for autonomous, multi-step scientific reasoning, and they hold immense potential for accelerating research in the physical sciences. However, comprehensive and in-depth evaluations of their capabilities within this domain remain lacking. To address this gap, we introduce PhySciBench, a benchmark highly relevant to physical science research, comprising 200 expert-curated questions, balanced between physics and chemistry, across six task categories that reflect real-world scientific workflows. Evaluations of state-of-the-art models and agent systems on PhySciBench reveal limited performance; even the strongest baseline, Gemini Deep Research, achieves an accuracy of only 33.5%. Analysis of failure cases identifies three recurrent deficiencies: fragility in extended reasoning chains, limited knowledge transfer across steps, and a lack of physics-grounded self-verification. Motivated by these findings, we develop DelveAgent, a modular multi-agent framework equipped with an adaptive planning loop, dual-granularity memory, and a hierarchical physics-grounded reflection mechanism. Across four scientific benchmarks, DelveAgent improves accuracy by up to 7.5 percentage points while reducing inference costs to approximately one-third of the strongest baseline. These results establish the significance of PhySciBench as a critical benchmark for evaluating AI systems in the physical sciences and demonstrate that architectural specialization can effectively enhance the reliability of autonomous scientific research.</p> - https://arxiv.org/abs/2606.18648 - Wed, 17 Jun 2026 00:00:00 +0000 - - - Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation - https://arxiv.org/abs/2606.18844 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.18844.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhilin Huang, Hang Gao, Ziqiang Dong, Yuan Chen, Yifeng Luo, Chujun Qin, Jingyi Wang, Yang Yang, Guanjun Jiang</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Self-distillation improves reasoning in large language models by using the model's own rollouts as training signal, typically through implicit logit-level alignment that minimizes KL divergence toward a privileged target distribution. However, because this supervision is generated via uncontrolled sampling, it provides no diagnostic insight into the model's specific errors or corrective guidance for its individual failure patterns. Consequently, the model learns to imitate a privileged distribution rather than receiving fine-grained corrections that pinpoint where and why its reasoning fails. In this paper, we propose Trajectory-Augmented Policy Optimization (TAPO), which advances self-distillation from implicit distributional alignment to explicit trajectory construction. During RL training, the model produces both correct and incorrect rollouts to the same query, and TAPO leverages this contrastive structure to construct micro-reflective corrections, new training trajectories that retain the model's erroneous reasoning up to the point of failure, then insert a natural-language diagnosis and corrected reasoning guided by a correct reference from the same sampling group. Since each trajectory is anchored in the learner's own prefix and solutions, the corrective signal preserves the model's on-policy distribution to a greater extent than the position-wise alignment imposed by KL-based methods. To integrate these trajectories, TAPO introduces difficulty-aware candidate selection at the model's capability boundary and decoupled advantage estimation to prevent gradient contamination. Experiments on AIME 2024, AIME 2025, and HMMT 2025 show that TAPO achieves consistent improvements over GRPO under the same number of training steps. Further analysis demonstrates that TAPO strengthens both first-pass reasoning and error-correction effectiveness.</p> - https://arxiv.org/abs/2606.18844 - Wed, 17 Jun 2026 00:00:00 +0000 - - - OpenRath: Session-Centered Runtime State for Agent Systems - https://arxiv.org/abs/2606.19409 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.19409.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Fukang Wen, Zhijie Wang, Ruilin Xu</p><p><b>Upvotes:</b> 68</p><p><b>Summary:</b> Modern agent systems often suffer from fragmented runtime state: transcripts, tool effects, memory events, workspace placement, branch provenance, and replay evidence are recorded separately and become difficult to inspect or reproduce. OpenRath addresses this issue with a PyTorch-like programming model for multi-agent, multi-session systems. The analogy concerns the role of a central first-class runtime abstraction, not tensor computation. Its core abstraction is Session, the runtime value passed between agents and workflows. A Session is branchable, inspectable, replayable, backend-aware, and composable. It records conversation chunks, sandbox placement, lineage metadata, token usage, pending work, and tool evidence, while defining where memory interactions enter the runtime record. Since this state is carried by the same value used in program execution, fork, merge, and replay become explicit runtime operations rather than states reconstructed from external traces. OpenRath further defines Sandbox, Tool, Agent, Memory, Workflow, and Selector, with Selector turning control flow into runtime-routed decisions. This report presents the programming model, architecture, audited milestones, and evidence protocol. Its claims are limited to controlled runtime properties, while broad quantitative comparisons, live-provider quality, optional-backend availability, and memory quality are left for follow-on evaluation. The central thesis is that Session provides agent systems with a first-class runtime value for auditable composition.</p> - https://arxiv.org/abs/2606.19409 - Wed, 17 Jun 2026 00:00:00 +0000 - - - Toward Parking Spot Occupancy Recognition: A Self-Supervised Approach - https://arxiv.org/abs/2606.20886 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20886.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Luan Marko Kujavski, Rayson Laroca, Paulo Lisboa de Almeida</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> As urban areas expand, automatic monitoring of parking lots becomes essential for efficient and sustainable cities. This work proposes a self-supervised approach for parking spot occupancy recognition that requires no labeled samples from the target parking lot. Building upon a self-supervised transfer learning fine-tuning protocol, the proposed training strategy consists of two self-supervised stages: first on unlabeled generic data and then on unlabeled target-specific data, followed by supervised fine-tuning using only generic parking lot labels. We adopt SimCLR with a ResNet-50 encoder and evaluate the method under a leave-one-out cross-environment protocol on three public datasets: PKLot, CNRPark-EXT, and PLds. We also introduce a two-stage deployment strategy in which a Strong General Model is initially deployed, followed by a Specialized Model that incorporates unlabeled images collected during the first N days of deployment in a self-supervised manner. Experimental results show that the Strong General Model alone outperforms supervised and self-supervised baselines, achieving an average accuracy of 97.2%, which further improves to 97.8% with the proposed two-stage strategy. These results demonstrate that self-supervised learning enables a scalable and labelefficient solution for real-world parking occupancy monitoring. Our trained models and source code are publicly available at https://github.com/LoanMaikon/Parking-Spot-Occupancy-Recognition.</p> - https://arxiv.org/abs/2606.20886 + EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policies + https://arxiv.org/abs/2606.20092 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20092.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ganlin Yang, Zhangzheng Tu, Yuqiang Yang, Sitong Mao, Junyi Dong, Tianxing Chen, Jiaqi Peng, Jing Xiong, Jiafei Cao, Jifeng Dai, Wengang Zhou, Yao Mu, Tai Wang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Memory remains a critical bottleneck for long-horizon robotic manipulation, as standard Vision-Language-Action (VLA) policies often fail when task-relevant cues become occluded or unobservable over time. While existing memory-augmented methods utilize historical context, they either suffer from severe information bottlenecks, incur high latency via decoupled dual systems, or rely on unselective buffers that accumulate massive visual redundancies. To address these limitations, we introduce EventVLA, an end-to-end framework founded on the concept of sparse visual evidence memory that comprises two core components: foundational visual anchors to retain initial and short-term contexts, and a dynamic Keyframe Evidence Memory (KEM) module. Specifically, KEM directly predicts future keyframe probabilities from the VLA's latent embeddings to autonomously capture and store sparse, task-critical visual events. This foresight-driven mechanism empowers the policy to dynamically evaluate the future causal utility of current observations, preserving transient visual evidence before it becomes unobservable. Furthermore, we propose RoboTwin-MeM, a diagnostic benchmark specifically designed to evaluate non-Markovian manipulation tasks with interactive visual evidence. Extensive evaluations show that across 17 memory-requiring simulation tasks and 4 real-world bimanual tasks, EventVLA achieves an average success rate improvement of +40% over state-of-the-art memory-augmented VLAs.</p> + https://arxiv.org/abs/2606.20092 Thu, 18 Jun 2026 00:00:00 +0000 - Robusto-2: Benchmarking Humans & VLMs for Autonomous Driving in Lima & New York City - https://arxiv.org/abs/2606.20980 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20980.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Adrian Cespedes, Marcelo Chincha, Dunant Cusipuma, Victor Flores-Benites, David Ortega, Arturo Deza</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> As Self-Driving Cars continue to expand internationally and use multi-modal systems such as VLMs as a cognitive backbone for their Action models; how well will these systems generalize in new settings, in particular out-of-distribution (OOD) edge-case scenarios in new geographies? In this paper, we study this open question by providing a full factorial analysis with human drivers of Lima, human drivers from New York City, and VLMs and showing them dashcam footage collected from Lima and New York City -- prompting them with a variety of questions under a Visual Question Answering (VQA) paradigm. In particular, we pick these two cities as they are highly challenging driving locations where no Self-Driving Car company currently operates in, and ask questions that span 4 categories: Factual, Ratings, Counterfactual and Reasoning. We find that Humans and VLMs diverge in their responses -- though this is modulated by the type of questions asked, and that Humans answer similarly independent of where they are from (Lima/NYC). To our surprise, we did not find a strong difference in terms of answers (Humans or VLMs) that was modulated by geography, likely due to their high out-of-distribution nature. Our dataset is available at: https://huggingface.co/datasets/Artificio/robusto-2</p> - https://arxiv.org/abs/2606.20980 + QG-MIL: A Gated Transformer Aggregator for Domain-Agnostic Multiple Instance Learning in Medical Imaging + https://arxiv.org/abs/2606.20027 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20027.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Luca Zedda, Davide Antonio Mura, Cecilia Di Ruberto, Maurizio Atzori, Muhammed Furkan Dasdelen, Carsten Marr, Andrea Loddo</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Attention-based Multiple Instance Learning aggregators in medical imaging are prone to attention concentration, producing overconfident and unstable predictions. We introduce QG-MIL, a gated transformer aggregator that addresses this through four synergistic architectural components: RMSNorm-based pre-normalization, per-head QK normalization, fine-grained attention output gating, and SwiGLU-style feed-forward modules. Together, these design choices stabilize training and distribute attention more uniformly across instances without auxiliary losses, masking, or multi-stage regularization. We evaluate QG-MIL across six benchmarks spanning whole-slide pathology and cell-level hematology, covering two fundamentally different MIL scales. The best-performing QG-MIL variants outperform leading baselines on all six benchmarks, with an average improvement of +6.1 mean macro F1 points. Attention overlays and attention mass analysis confirm more distributed instance weighting. Ablation studies show that while individual components can match the full model on specific datasets, the QG-MIL design provides the most consistent cross-domain performance and tightest variance when compared to selected baselines. We release a configurable implementation to support reproducibility at: https://github.com/unica-visual-intelligence-lab/QG-MIL</p> + https://arxiv.org/abs/2606.20027 Thu, 18 Jun 2026 00:00:00 +0000 - Go-with-the-Track: Video Compositing and Motion Control with Point Tracking - https://arxiv.org/abs/2606.20891 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20891.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Koichi Namekata, Yash Kant, Zhizheng Liu, Ryan D Burgert, Yuancheng Xu, Kuan Heng Lin, Emmett Steven, Julien Philip, Li Ma, Andrea Vedaldi, Paul Debevec, Ning Yu</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Filmmaking demands precise motion control and reference image compositing -- capabilities that existing methods treat separately. Point-track-conditioned image-to-video models restrict content insertion to the first frame, while reference-to-video models lack fine-grained spatial-temporal control over how reference content integrates across frames. We present Go-with-the-Track, which unifies both capabilities by jointly conditioning on multiple reference images and reference-anchored point-tracks -- extending conventional point-tracks to explicitly establish correspondences between generated frames and reference images, thus enabling precise compositing and motion control throughout the video. To achieve this, we introduce spatially-aware point-track embeddings that encode the full sequence of point-track coordinates using a coordinate-wise MLP followed by temporal pooling. This representation captures the spatial characteristics of each point-track (serving as a unique identifier), while the embedding similarity correlates directly with spatial proximity, enhancing the model's ability to distinguish and associate point-tracks. We inject these point-track embeddings into a video diffusion transformer via a lightweight adapter, resolving the pixel-to-patch resolution mismatch while avoiding the substantial motion detail loss inherent in naive point-track subsampling. We use a hybrid training strategy to train jointly on dynamic, static, and synthetic scene video datasets to boost motion controllability. Experiments demonstrate that Go-with-the-Track achieves superior motion and reference control in a single model and enables new capabilities: multi-reference conditioned video generation with point-track driven compositing, as well as camera control for both static and dynamic scenes. Project Page: https://eyeline-labs.github.io/Go-with-the-Track/</p> - https://arxiv.org/abs/2606.20891 + MemGUI-Agent: An End-to-End Long-Horizon Mobile GUI Agent with Proactive Context Management + https://arxiv.org/abs/2606.19926 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.19926.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Guangyi Liu, Gao Wu, Congxiao Liu, Pengxiang Zhao, Liang Liu, Mading Li, Qi Zhang, Mengyan Wang, Liang Guo, Yong Liu</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> MLLM-based mobile GUI agents have made substantial progress on short-horizon tasks, yet remain unreliable on long-horizon tasks that require retaining intermediate facts across many steps and app transitions. We attribute this limitation to ReAct-style prompting, which passively accumulates per-step records, leading to prompt explosion and dilution of critical cross-app facts. To address this, we introduce MemGUI-Agent, an end-to-end long-horizon mobile GUI agent with proactive context management. MemGUI-Agent is built on Context-as-Action (ConAct), which casts context management as first-class actions emitted by the same policy that selects UI actions. Instead of passively appending history, ConAct maintains three structured context fields: folded action history, folded UI state, and recent step record, preserving critical UI facts while keeping context compact. To make proactive context management learnable across model scales, we construct MemGUI-3K, a 2,956-trajectory dataset with full ConAct annotations for supervised training and offline analysis. Training an 8B model on MemGUI-3K produces MemGUI-8B-SFT, an 8B MemGUI-Agent that achieves the best open-data 8B performance on MemGUI-Bench and generalizes to the out-of-distribution MobileWorld benchmark. Code, data, and trained models will be released at https://memgui-agent.github.io/.</p> + https://arxiv.org/abs/2606.19926 Thu, 18 Jun 2026 00:00:00 +0000 - Manifold Bandits: Bayesian Curriculum Learning over the Latent Geometry of Large Language Models - https://arxiv.org/abs/2606.19750 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.19750.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Darrien McKenzie, Nicklas Hansen, Xiaolong Wang</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Reinforcement learning (RL) is a central approach for improving reasoning capabilities in large language models (LLMs), where training efficiency depends critically on how problems are sampled during optimization. Existing adaptive curriculum learning methods typically prioritize prompts of intermediate difficulty, treating problem selection as a standard bandit problem with independent arms and overlooking the structured, heterogeneous nature of the task space. In this work, we frame problem sampling as a manifold-structured bandit problem with endogenous non-stationarity: problems are related through the model's latent representation space, and sampling decisions can steer how learning signals evolve across that space. To operationalize this perspective, we introduce Bayesian Manifold Curriculum (BMC), a structure-aware framework that organizes problems into a hierarchical task tree and applies Bayesian learning to guide sampling. Empirically, we find that different sampling strategies induce non-trivial tradeoffs between productivity (learning signal), diversity (coverage of the task manifold), and utility (evaluation relevance). These results show that prioritizing difficulty alone is insufficient for strong downstream performance, highlighting the importance of incorporating structure and type-awareness into problem sampling.</p> - https://arxiv.org/abs/2606.19750 + MobileForge: Annotation-Free Adaptation for Mobile GUI Agents with Hierarchical Feedback-Guided Policy Optimization + https://arxiv.org/abs/2606.19930 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.19930.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Guangyi Liu, Pengxiang Zhao, Gao Wu, Yiwen Yin, Mading Li, Liang Liu, Congxiao Liu, Zhang Qi, Mengyan Wang, Liang Guo, Yong Liu</p><p><b>Upvotes:</b> 34</p><p><b>Summary:</b> MLLM-based mobile GUI agents have made substantial progress in UI understanding and action execution, but adapting them to real target apps remains costly because mobile apps are numerous, frequently updated, and hard to cover with human-written tasks, demonstrations, or reward labels. Existing annotation-free GUI learning reduces manual supervision, yet lacks a unified substrate connecting target-app exploration, curriculum mining, rollout execution, and feedback, while policy optimization often relies on isolated rollouts and coarse rewards that are hard to convert into reliable improvement signals. We present MobileForge, an annotation-free adaptation system for mobile GUI agents. MobileForge consists of MobileGym, which grounds task generation and rollout evaluation in real mobile app interaction, and Hierarchical Feedback-Guided Policy Optimization (HiFPO), which turns trajectory outcomes, step-level process feedback, and corrective hints into hint-contextualized step-level GRPO updates. Using only automatically generated annotation-free adaptation data, MobileForge adapts Qwen3-VL-8B to 67.2% Pass@3 on AndroidWorld, close to the closed-data GUI-specialized GUI-Owl-1.5-8B base model at 69.0%. The MobileForge-adapted ForgeOwl-8B further reaches 77.6% Pass@3 on AndroidWorld and 41.0% success on the out-of-domain MobileWorld GUI-only split, establishing the strongest open-data mobile GUI agent in our evaluation. Code, data, and trained models will be released at https://mobile-forge.github.io/.</p> + https://arxiv.org/abs/2606.19930 Thu, 18 Jun 2026 00:00:00 +0000 - Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning - https://arxiv.org/abs/2606.20002 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20002.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yanxi Chen, Weijie Shi, Yuexiang Xie, Boyi Hu, Yaliang Li, Bolin Ding, Jingren Zhou</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> This work presents a general framework for training large language models (LLMs) to "Connect the Dots" (CoD), a meta-capability required by long-lifecycle agents: as an LLM-based AI agent gets deployed in an environment, it solves a long sequence of tasks while continuously exploring the environment, learning from its own experiences, and iteratively self-updating its context about the environment, thereby achieving progressively better performance on future tasks conditioned on the updated context. Major components of the CoD framework include: (1) algorithm design and infrastructure for end-to-end reinforcement learning (RL) with long rollout sequences interleaving solve-task and update-context episodes; (2) tasks and environments for incentivizing and eliciting the targeted meta-capability in LLMs during training, as well as for faithfully measuring progress during evaluation. We present proof-of-concept implementations of the CoD framework, including a GRPO-style RL algorithm with fine-grained credit assignment, as well as tasks and environments tailored to the targeted meta-capability (rather than domain-specific LLM capabilities or standard task-by-task RL). Empirical results validate the efficacy of end-to-end RL training in the CoD setting, and demonstrate the potential for out-of-distribution generalization -- within the training domains, across different domains, and from CoD to Ralph-loop settings -- of the elicited meta-capability. Our investigation of CoD connects several lines of prior works, and opens up new opportunities for advancing LLMs and AI agents. To facilitate further research and applications, we release our implementations at https://github.com/agentscope-ai/Trinity-RFT/tree/research/cod/examples/research_cod.</p> - https://arxiv.org/abs/2606.20002 - Thu, 18 Jun 2026 00:00:00 +0000 - - - HydraHead: From Head-Level Functional Heterogeneity to Specialized Attention Hybridization - https://arxiv.org/abs/2606.20097 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20097.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhentao Tan, Wei Chen, Jingyi Shen, Yao Liu, Xu Shen, Yue Wu, Jieping Ye</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> The quadratic complexity of attention poses a critical bottleneck for long-context processing, spurring interest in hybrid attention designs. Most open-source hybrid models adopt a layer-wise strategy. Yet, prior work has noted the inherent difficulty of integrating Linear Attention (LA) with Full Attention (FA), suggesting that the design space of attention hybridization remains underexplored. To probe this space, we conduct interpretability analysis and observe that layers exhibit block-wise functional similarity, while individual heads within the same layer display distinct functional specialization despite sharing input features. This head-level heterogeneity suggests that the head dimension provides a natural and principled granularity for fusing heterogeneous attention signals. Building on this insight, we introduce HydraHead, a novel architecture that hybridizes FA and LA along the head axis. HydraHead features two key innovations: (1) an interpretability-driven selection strategy that identifies retrieval-critical heads and preserves FA only for them, and (2) a scale-normalized fusion module that reconciles the distributional gap between FA and LA head outputs. By leveraging a three-stage transfer pipeline with parameter reuse and distillation, we achieve high-performance hybrid models with minimal training overhead. Under a unified training setup, HydraHead outperforms other hybrid designs in long-context tasks while maintaining strong general reasoning. With interpretability-driven head selection, it matches a 3:1 layer-wise hybrid's long-context performance at a 7:1 LA-to-FA ratio. Crucially, trained on only 15B tokens, HydraHead achieves over 69% improvement over the baseline at 512K context length, approaching Qwen3.5, a leading model of comparable size with a native context length of 256K. This highlights the significant scaling potential of head-level hybridization.</p> - https://arxiv.org/abs/2606.20097 - Thu, 18 Jun 2026 00:00:00 +0000 - - - World Action Models: A Survey - https://arxiv.org/abs/2606.20781 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20781.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qiuhong Shen, Shihua Zhang, Yue Liao, Qi Li, Zhenxiong Tan, Shizun Wang, Shuicheng Yan, Xinchao Wang</p><p><b>Upvotes:</b> 37</p><p><b>Summary:</b> World Action Models (WAMs) are embodied predictive-action models that make a forecast of the future available to action. Recent WAMs repurpose large video generation models, and a parallel line relies on language or vision-language backbones without a video-generation core. This rapid expansion has blurred the boundary among broad world models, video generation models, action-grounded video world models, Vision-Language-Action policies, and WAMs. This survey gives the field a common account. It first clarifies these boundaries, then organizes existing works through two complementary views. The first view asks what each method is required to generate, spanning rendered futures, latent futures, and video-generation-free action reasoning. The second view decomposes each method by predictive substrate, backbone, action coupling, and deployment regime. This anatomy supports a unified discussion of interactability, causality, persistence, physical plausibility, and generalization, followed by data, evaluation, and open challenges. Across these axes, a consistent design pattern emerges: WAMs are not simply video generators with action heads, but predictive-action methods whose design choices trade representational richness against compute, memory, latency, and action-label cost. The field is moving toward methods that generate less of the future while preserving what control requires. The survey homepage is available at https://world-action-models.github.io/.</p> - https://arxiv.org/abs/2606.20781 - Thu, 18 Jun 2026 00:00:00 +0000 - - - Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention - https://arxiv.org/abs/2606.20945 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.20945.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Vishesh Tripathi, Abhay Kumar</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> Self-attention is central to Transformer performance and is often the most expensive part of the Transformer at long context lengths because its pairwise token interactions scale quadratically with sequence length. Standard dense attention also applies the same set of attention heads to every token regardless of token difficulty or information content. This uniform activation can waste compute, especially as sequences grow longer and attention cost increases rapidly. We propose Grouped Query Experts (GQE), a mixture-of-experts layer on top of grouped-query attention (GQA). Within each GQA group, a router selects k query-head experts per token while all key-value (KV) heads remain dense and unchanged. Thus, GQE keeps the KV cache benefits of GQA and reduces only the active query-head computation. On a fixed 30B token budget at the 250M parameter scale, GQE matches the all-active GQA baseline in downstream accuracy while activating half the query heads per token.</p> - https://arxiv.org/abs/2606.20945 - Thu, 18 Jun 2026 00:00:00 +0000 - - - Improving Text-to-Music Generation with Human Preference Rewards - https://arxiv.org/abs/2606.21670 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21670.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yonghyun Kim, Junwon Lee, Haiwen Xia, Yinghao Ma, Chris Donahue</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> We describe our entry to the efficiency track of the Academic Text-to-Music (ATTM) Grand Challenge at ICME 2026. Beyond the challenge protocol's FAD-CLAP and CLAP score, we add a learned human-preference reward from TuneJury, a twin pairwise ranker trained over open music-preference datasets. The reward serves both as a training-time conditioning signal and as a sample-selection criterion. The pipeline combines five engineering decisions on a 120M-parameter FluxAudio-S backbone, four at training time and one at inference: (i) training-time reward conditioning that doubles as an inference-time CFG axis, (ii) a sweep over five score-conditioning architectures, where training and inference use different variants, (iii) expert iteration on the top decile, (iv) a short preference-tuning pass (CRPO) for audio-text alignment, and (v) inference post-processing via joint CFG, source separation, and loudness normalization. Per-stage decomposition on 100 Song Describer prompts shows training-time reward conditioning as a functional conditioning axis, expert iteration as the dominant contributor, the preference-tuning pass adding only noise-level gain, and the inference-time score scalar already saturated by the end of the chain.</p> - https://arxiv.org/abs/2606.21670 - Fri, 19 Jun 2026 00:00:00 +0000 - - - An Exploratory Case Study of LLM-Assisted Refactoring and Gameplay Feature Generation in an Endless Runner Game - https://arxiv.org/abs/2606.21171 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21171.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jan Wunderlich, Markus Kleffmann, Sebastian Lempert</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Large language models (LLMs) are increasingly used to support software development, but their practical usefulness in applied game-development settings remains underexplored, especially when generated code must be integrated into an existing game software system. This paper presents an exploratory empirical case study of GPT-4o in a custom Python/Pygame endless runner. The study examines six selected development tasks: three localized refactoring tasks and three tasks involving gameplay feature generation. The resulting implementations were evaluated using software metrics, unit tests, and manual gameplay assessments. In this case study, all three selected refactoring tasks were completed successfully in functional terms, whereas only one of the three selected gameplay feature generation tasks resulted in a correctly integrated feature. The findings suggest that, in this setting, GPT-4o handled localized transformations more reliably than tasks requiring new gameplay interactions across multiple existing systems. Given the exploratory single-case design, these results are best interpreted as indicative observations rather than as generalizable evidence of category-level model performance. Overall, the paper contributes a transparent case-based account of the opportunities and limitations of LLM-assisted refactoring and gameplay feature generation in an existing game software system.</p> - https://arxiv.org/abs/2606.21171 - Fri, 19 Jun 2026 00:00:00 +0000 - - - CalVerT: Augmenting Agents with Calibrated Verifier Telemetry Improves Action and Learning in Knowledge-Intensive Tasks - https://arxiv.org/abs/2606.21777 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21777.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ashwin Vinod, Ying Ding, Elias Stengel-Eskin</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> LLM agents in knowledge intensive question answering take retrieval and reasoning actions with incomplete knowledge about whether their current answer is uncertain, unsupported, or already complete. This produces two failure modes: committing to confident but unsupported answers, which hurts accuracy, and over-retrieving when the evidence in hand already suffices, resulting in wasted compute. To give agents a more complete picture of the state space they are operating in, we introduce calibrated verifier telemetry (CalVerT), which augments the agent's state with additional telemetry: a calibrated self-confidence score and a grounding verifier score. We show that CalVerT can improve agents in both training-free and training-based settings. On four QA benchmarks, we find that CalVerT raises F1 by triggering retrieval in cases where agents over-rely on parametric knowledge, while cutting redundant retrieval in cases where agents have sufficient context to answer. We show that CalVerT can augment existing QA frameworks without training. Moreover, CalVerT also improves trained systems: by simply augmenting an agent's state with telemetry, we observe improvements after reinforcement learning, as compared to an agent with identical training but no CalVerT telemetry.</p> - https://arxiv.org/abs/2606.21777 - Fri, 19 Jun 2026 00:00:00 +0000 - - - Demystifying Training-Time Augmentation for Data-Constrained Language Model Pretraining - https://arxiv.org/abs/2606.16246 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.16246.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Michael K. Chen, Xikun Zhang, Fan Bai, Zhengding Hu, Zhen Wang</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> As AI labs approach a data ceiling where compute capacity outpaces the rate of new high-quality text generation, language model pretraining is shifting toward a data-constrained, compute-abundant regime that demands productive multi-epoch training on fixed corpora. Standard autoregressive (AR) pretraining overfits severely in this setting, reaching its optimum early and then continuously deteriorating. We investigate training-time data augmentation as a regularizer to mitigate this overfitting and enable productive training for hundreds of epochs on the same data. We introduce three orthogonal categories of augmentation for AR pretraining: token-level noise (masking, random replacement), sequence permutations (right-to-left prediction, Fill-in-the-Middle), and target offset prediction (x_{t+i} for i > 1). Through systematic ablations, we find that individual augmentations delay overfitting and lower validation loss relative to the baseline, with random token replacement achieving the best minimum loss among individual methods. Combining augmentation categories further lowers the minimum validation loss. Our experiments demonstrate that data augmentations mitigate AR pretraining's data inefficiency and offer a promising solution to the data-constrained regime~\footnote{All code and data are available at https://github.com/ michaelchen-lab/ data-augmentations-for-pretraining.</p> - https://arxiv.org/abs/2606.16246 - Fri, 19 Jun 2026 00:00:00 +0000 - - - Counsel: A Meta-Evaluation Dataset for Agentic Tasks - https://arxiv.org/abs/2606.21627 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21627.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sashank Pisupati, Henry Broomfield, Eujeong Choi, Antonia Calvi, Charlie Wang, Roman Engeler, Max Bartolo, Patrick Lewis</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> As agentic systems tackle increasingly complex multi-step tasks, evaluating their trajectories presents a major bottleneck - human annotation of a single trajectory on popular agentic benchmarks can take hours, making it difficult to scale evaluations for measuring performance or curating training data. This has driven widespread reliance on automated approaches such as LLM-as-a-judge (LLMJ) to critique agents at the process and outcome-levels at scale, however, the soundness of LLMJ critiques often goes unmeasured. Here, we introduce Counsel, the first public dataset of meta-evaluations for agentic tasks. Counsel consists of process-level critiques from open-weight LLMJs on two agent benchmarks: tau-bench (customer support agents) and DA-Code (coding agents), and human meta-evaluations of these critiques. Human annotators label critiques on each flagged error as "spot on", "correct location but poor reasoning", or "should not have flagged", achieving reliable inter-annotator agreement (Krippendorff's alpha of 0.78). The resulting dataset stratifies LLMJ critiques by human alignment across both error location within a trajectory and reasoning quality, serving as valuable data to calibrate, improve, or train LLMJs for agents. Comparing open-weight judges, we find that more capable judge models and more reasoning effort both enabled improved human agreement, with the strongest judge reaching ~88% agreement on location and ~65% on reasoning. Counsel is generated using open-weight models and is permissively licensed for broad community use, which we hope will enable rigorous study and improved alignment of LLM-based evaluators for agentic systems.</p> - https://arxiv.org/abs/2606.21627 - Fri, 19 Jun 2026 00:00:00 +0000 - - - PoLAR: Factorizing Extent and Mode in Latent Actions for Robot Policy Learning - https://arxiv.org/abs/2606.21139 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21139.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Youngjoon Jeong, Jihwan Yu, Minsoo Jo, Junha Chun, Taesup Kim</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Latent action pretraining learns representations of visual change from pairs of observations, but existing methods typically encode each transition as a single unstructured representation that entangles transition extent and transition mode. We introduce Polar Latent Actions with Radial structure (PoLAR), which imposes a radial-direction structure on latent actions, encouraging radius to encode transition extent and direction to retain transition mode. PoLAR uses temporal offset between two observations as a weak proxy for transition extent, encouraging latent action from observation pairs separated by larger temporal gaps to occupy larger radii. We instantiate this structure in hyperbolic space, whose expanding volume with radius offers a natural fit for more diverse transition modes at larger extents. Across in-task and large-scale pretraining settings, PoLAR improves downstream policy performance in simulation and real-world robot experiments, outperforming latent action baselines and strong pretrained VLAs. These results suggest that the geometry of the latent action space is an important design choice for transferring visual pretraining to downstream robot policy learning.</p> - https://arxiv.org/abs/2606.21139 - Fri, 19 Jun 2026 00:00:00 +0000 - - - Toward Open Weight Models Without Risks: Separating Public and Private Capabilities in LLMs - https://arxiv.org/abs/2606.21638 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21638.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Charbel El Feghali, Arkil Patel, Nicholas Meade, Spandana Gella, Verna Dankers, Siva Reddy</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Open-weight Large Language Models (LLMs) enable scientific progress and broad deployment. However, they make it difficult to control access to sensitive capabilities. Current practice either suppresses dangerous capabilities before release or mediates access through closed services that use specialized model variants, input/output monitors, and API permissions. The former is susceptible to jailbreaks while sacrificing capability for all users to mitigate the risks posed by a few, and the latter is fundamentally incompatible with open-weight release. In this paper, we propose Tiered Language Models (TLMs), where a single set of released weights supports multiple capability levels. In its default public configuration, a TLM behaves as a conventional LLM. A compact secret key specifies a permutation over a small parameter subset, inducing an alternative computation graph over the same weights that exposes additional capabilities. We develop a training protocol that jointly pretrains both configurations from scratch, then fine-tunes the keyed configuration on private data with regularization to preserve the public model's behavior. We pretrain 180M- and 650M-parameter TLMs and demonstrate that the keyed configuration can acquire a new language, gain instruction-following ability, and memorize private factual knowledge, whereas the public configuration exhibits none of these capabilities. Moreover, we show that our approach extends naturally to multiple hierarchical tiers. Because authorization operates on the model's weight structure rather than in the input space, the mechanism resists fine-tuning-based extraction and partial key compromise. In general, TLMs take a step toward reconciling open-weight release with selective capability control.</p> - https://arxiv.org/abs/2606.21638 - Fri, 19 Jun 2026 00:00:00 +0000 - - - EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory - https://arxiv.org/abs/2606.21649 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21649.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chang Nie, Chaoyou Fu, Junlan Feng, Caifeng Shan</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> Existing embedding models are inherently static: they encode text segments in isolation, ignoring their surrounding context and temporal order. This paper introduces EvoEmbedding, a novel embedding model that generates evolvable representations for retrieval. It is tailored for long-context scenarios, where information is dynamic, sequential, and requires continuous state tracking. Our design is simple: EvoEmbedding maintains a continuously updated latent memory as it sequentially processes inputs, and uses it alongside the raw content to jointly generate evolvable embeddings. Consequently, for the same query, our model adapts its representation to retrieve distinct targets based on the evolving context, going beyond static semantic search. To equip the model with this capability, we construct EvoTrain-180K, a diverse dataset for the joint optimization of latent memory and retrieval. Furthermore, we introduce a memory queue to prevent representation collapse during recurrent encoding, alongside segment-batching techniques that tackle significant length variance and accelerate training by 3.8times. Extensive experiments show that our model not only outperforms larger-scale specialists (e.g., Qwen3-Embedding-8B and KaLM-Embedding-Gemma3-12B) across a range of long-context retrieval benchmarks, but also generalizes well to downstream tasks (e.g., personalization) with contexts 10times longer than its training window. Notably, EvoEmbedding seamlessly integrates into agentic workflows to boost performance. For instance, a naive RAG pipeline equipped with our model surpasses dedicated agentic memory systems. Project Page: https://clare-nie.github.io/EvoEmbedding.</p> - https://arxiv.org/abs/2606.21649 - Fri, 19 Jun 2026 00:00:00 +0000 - - - DataClaw0: Agentic Tailoring Multimodal Data from Raw Streams - https://arxiv.org/abs/2606.21337 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21337.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Cong Wan, Zeyu Guo, Zijian Cai, Jiangyang Li, SongLin Dong, Lin Peng, Xiangyang Luo, Zhiheng Ma, Yihong Gong</p><p><b>Upvotes:</b> 65</p><p><b>Summary:</b> Massive unstructured multimodal streams suffer from high "data entropy," impeding both efficient human knowledge acquisition and high-quality AI post-training. Existing passive annotation paradigms, heavily reliant on heuristic rules or general VLMs, are costly, monotonous, and fail to unlock the deep procedural logic embedded in raw data. We elevate data processing to a learnable capability, proposing a paradigm shift towards Agentic Data Tailoring, which actively refining and structuring data to align with diverse user and downstream intents. To overcome the data scarcity bottleneck in training such high-order capabilities, we design a two-stage pipeline grounding generative semantic synthesis in deterministic Factual Anchors, yielding a large-scale dataset spanning five core physical and digital domains. Building upon this, DataClaw_0-9B model synergizes Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO), achieving robust alignment with complex refinement and tailoring intents. To systematically quantify this capability, we construct DataClaw_0-val, the first benchmark dedicated to data refinement. Crucially, we adopt downstream post-training as the ultimate validation touchstone. Evaluations on video generation, real-world VQA, and GUI navigation confirm that DataClaw_0 delivers high-information-density tailored data, facilitating efficient model adaptation to new tasks under limited training data regimes. Project page: https://czjdsg.github.io/MakeAnyData</p> - https://arxiv.org/abs/2606.21337 - Fri, 19 Jun 2026 00:00:00 +0000 - - - A Verifiable Search Is Not a Learnable Chain-of-Thought - https://arxiv.org/abs/2606.21884 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21884.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Harsh Patel</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> It is tempting to assume any task solvable by a short program can be taught to a model as its chain-of-thought: write the steps out, fine-tune, and the model follows. This paper shows the assumption fails for an identifiable class of procedures. The testbed is nine reasoning tasks, each from a deterministic generator; public and hidden splits share generators, so held-out data proxies test accuracy. I reverse-engineer the generators into Python solvers, render them as chain-of-thought, and distill into a rank-<= 32 LoRA over a 30B (3.5B-active) Nemotron model. Forward-computable tasks install readily: lookup/arithmetic and an 8-bit boolean task transfer (>= 0.99 and 0.68). Cryptarithm does not: distilling its backtracking search holds at 0.01-0.07 across eleven chain-of-thought designs, RL from verifiable rewards, and self-training, even though a search solver answers 71% of instances. This is not a capability gap. The model does the arithmetic on 97-100% of lines and ranks the correct cipher in its top eight on 71%; it cannot carry the search forward as a left-to-right derivation. Fine-tuning learns the shape of a verifiable elimination step while its verdicts become unconditional templates, correct only 16-57% of the time ("verdict-as-token"). The ceiling holds across backbones from 3B to 671B and across fine-tuning and prompting; a controlled intervention isolates the cause: revealing the cipher key, which turns the derivation forward, lifts the same instances from 0.03 to 0.57. When a procedure's only solution is search over information-free structure, no faithful forward chain-of-thought exists to imitate. The task becomes learnable only by removing the search, precomputing its combinatorial core into a catalog and reducing the trace to recall plus verification; the 1st-place solution reaches Private LB 0.92 this way. What distills is memorization and verification, not search.</p> - https://arxiv.org/abs/2606.21884 + Multi4D: High-Fidelity Dynamic Gaussian Splatting via Multi-Level Competitive Allocation + https://arxiv.org/abs/2606.22197 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22197.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Rui Wang, Quentin Lohmeyer, Siyu Tang, Mirko Meboldt</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Dynamic 3D Gaussian splatting faces a fundamental tension between motion consistency and visual fidelity. Deformation-based approaches preserve temporal correspondence but suffer from motion over-factorization, oversmoothing high-frequency dynamics. In contrast, 4D-primitive methods capture fine visual details yet incur temporal overparameterization, breaking object identity and leading to severe storage overhead. To resolve this, we introduce Multi4D, a framework for high-fidelity dynamic Gaussian Splatting based on multi-level competitive allocation. Instead of a monolithic representation, we distribute modeling capacity across three structured levels: static structure, persistent dynamic geometry, and transient appearance primitives. Through shared rasterization and residual-driven optimization, these levels dynamically compete to explain photometric error, enabling adaptive specialization without pre-assigned decomposition. This allocation preserves long-term motion consistency while capturing fine dynamic detail, achieving state-of-the-art rendering quality and real-time performance with significantly fewer dynamic primitives. Furthermore, because our representation explicitly tracks compact persistent Gaussians over time, semantic features can be embedded afterward, enabling Multi4D to achieve state-of-the-art 4D segmentation accuracy with an order-of-magnitude speedup. Project page: https://batfacewayne.github.io/Multi4D.io/</p> + https://arxiv.org/abs/2606.22197 Sat, 20 Jun 2026 00:00:00 +0000 - Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding - https://arxiv.org/abs/2606.21906 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.21906.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xuanming Zhang, Sining Zhoubian, Yuxuan Chen, Tianyi Tang, An Yang, Sean Du, Chujie Zheng, Fei Huang, Dayiheng Liu, Gao Huang, Jingren Zhou</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Autoregressive generation in large language models (LLMs) conventionally decodes from the final layer, assuming that deeper representations yield more reliable next-token predictions. We revisit this assumption by revealing a recurring Guess-Refine-Perturb dynamic: early layers form coarse guesses, intermediate layers refine reasoning-relevant semantics, and final layers can perturb these refined predictions toward generic or alignment-preferred tokens. We introduce Confident Decoding, a training-free decoding strategy that dynamically selects the most reliable near-final layer through entropy-guided conservative backward search. We further provide a theoretical formulation of layer selection as an optimal stopping problem, showing that under bounded projection noise and dominant late-stage alignment perturbation, our search rule filters perturbation while bounding the loss relative to the oracle refinement layer. Experiments across dense and Mixture-of-Experts LLMs demonstrate consistent gains on challenging reasoning benchmarks, including GPQA-Diamond, Omni-MATH, and HLE, with zero memory overhead and less than 2% latency increase. These results suggest dynamically bypassing final-layer perturbations can unlock stronger reasoning behavior from aligned LLMs.</p> - https://arxiv.org/abs/2606.21906 - Sat, 20 Jun 2026 00:00:00 +0000 - - - BioMatrix: Towards a Comprehensive Biological Foundation Model Spanning the Modality Matrix of Sequences, Structures, and Language - https://arxiv.org/abs/2606.22138 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22138.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qizhi Pei, Zhimeng Zhou, Yi Duan, Yiyang Zhao, Wei Li, Han Guo, Liang He, Chengping Li, Chang-Yu Hsieh, Conghui He, Rui Yan, Lijun Wu</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> We present BioMatrix, the first multimodal foundation model that natively integrates sequences, structures, and natural language for both molecules and proteins within a single decoder-only architecture. Existing biological foundation models pursue native multimodality and broad entity coverage separately: those that fuse multiple modalities under a shared objective remain confined to a single entity type, while those spanning multiple entity types either omit explicit structural modeling or rely on adapter-based designs in which the model cannot natively generate the very modalities it can read. BioMatrix closes this gap by mapping molecular sequences (supporting both SMILES and SELFIES notations), molecular structures, protein sequences, protein structures, and natural language into a shared discrete token space through a unified tokenization scheme, so that all modalities are consumed and produced uniformly under a single next-token prediction objective -- without external encoders, projection adapters, or modality-specific output heads. Built upon the Qwen3 language model (1.7B and 4B), BioMatrix is continually pretrained on 304.4 billion tokens spanning general and domain-specific text, sequence and structure views of molecules and proteins, and cross-modal corpora that interleave biomolecular entities with scientific text and link distinct entities through molecule-protein and protein-protein interaction data. After tuning on a comprehensive suite of downstream applications covering 80 tasks across 6 categories -- encompassing single-entity and multi-entity understanding and generation tasks across and within modalities -- BioMatrix achieves state-of-the-art or competitive performance on 77 out of 80 tasks, demonstrating that a single, natively multimodal generalist model can effectively match or surpass specialized approaches across a wide range of biological tasks.</p> - https://arxiv.org/abs/2606.22138 - Sat, 20 Jun 2026 00:00:00 +0000 - - - Libretto: Giving LLM Agents a Sense of Musical Structure - https://arxiv.org/abs/2606.22708 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22708.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yichen Xu</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Generative music systems can now produce impressive audio from text prompts, but audio outputs are difficult to inspect, edit, and diagnose as musical structure. We introduce Libretto, an agent-facing framework for symbolic music generation and revision. Libretto uses an LLM-native grammar with explicit onset slots, voices, and bar-level organization, then evaluates each piece in a corpus-calibrated statistical space over rhythm, harmony, melody, texture, form, and variation. The same structural axes support retrieval, diagnosis, copy-risk control, and iterative self-revision. Across gap filling, reference-guided full-piece generation, gradual morphing, and educational music generation, Libretto turns symbolic music from a raw token sequence into a measurable and editable object for language-model agents.</p> - https://arxiv.org/abs/2606.22708 - Sun, 21 Jun 2026 00:00:00 +0000 - - - PolicyTrim: Boosting Intrinsic Policy Efficiency of Vision-Language-Action Models - https://arxiv.org/abs/2606.22540 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22540.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xianghui Wang, Feng Chen, Wenbo Zhang, Hua Yan, Zixuan Wang, Changsheng Li, Yinjie Lei</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Vision-Language-Action (VLA) models provide a unified paradigm for robotic manipulation, yet their real-world deployment is often bottlenecked by execution efficiency. While existing efforts predominantly focus on compute-centric efficiency to reduce per-step inference latency, the intrinsic policy efficiency of these models remains largely unexplored. Policy efficiency is fundamentally affected by two factors, namely the effective executable length of predicted action chunks and the total physical steps required to complete a task. These two factors jointly determine the total number of forward inference calls during execution. We observe that current VLA policies struggle with planning unreliability and action redundancy, suffering from severe prediction degradation at the tail of action chunks and tending to generate unnecessarily redundant physical steps. To address this, we propose PolicyTrim, a reinforcement learning-based post-training framework that extends the reliable action chunk length and reduces redundant physical steps. For reliable chunk extension, we employ a dynamic exploration strategy that explicitly rewards the successful completion of longer executable lengths, progressively pushing the trustworthy prediction horizon to its empirical limit. For step efficiency, we design a redundancy-aware reward that directly favors successful task completions with fewer steps while penalizing unreproducible shortcuts, effectively eliminating redundant physical actions. Extensive experiments across three benchmarks and three VLA models demonstrate that PolicyTrim improves action chunk utilization by 3times and reduces physical execution steps by 51.4\%. Ultimately, our framework delivers up to a 5.83times end-to-end deployment speedup without compromising task success rates.</p> - https://arxiv.org/abs/2606.22540 - Sun, 21 Jun 2026 00:00:00 +0000 - - - PlanBench-XL: Evaluating Long-Horizon Planning of LLM Tool-Use Agents in Large-Scale Tool Ecosystems - https://arxiv.org/abs/2606.22388 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22388.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiayu Liu, Qihan Lin, Cheng Qian, Rui Wang, Emre Can Acikgoz, Xiaocheng Yang, Jiateng Liu, Zhenhailong Wang, Xiusi Chen, Heng Ji, Dilek Hakkani-Tür</p><p><b>Upvotes:</b> 78</p><p><b>Summary:</b> LLM agents increasingly operate in large tool ecosystems, where real-world tasks require discovering relevant tools, inferring implicit sub-goals, and adapting to dynamic environments over long horizons. However, existing benchmarks rarely evaluate planning under retrieval-limited tool visibility. To address this gap, we introduce PlanBench-XL, an interactive benchmark of 327 retail tasks over 1,665 tools that tests whether agents can iteratively retrieve usable tools, invoke them to uncover intermediate evidence for subsequent calls toward the final goal. PlanBench-XL further features an optional blocking mechanism that simulates real-world unpredictability through missing, failing, or distracting tool functions, forcing agents to detect disrupted paths and adapt at runtime. Experiments on ten leading LLMs show that massive-tool planning remains challenging: while GPT-5.4 achieves 51.90% accuracy in block-free settings, it collapses to 11.36% under the most severe blocking condition. Further analysis shows that agents are especially vulnerable when failures lack explicit error signals or when recovery requires longer alternative tool-use paths. These results establish PlanBench-XL as a testbed for diagnosing agentic planning failures and highlight the need for robust adaptive planning in long-horizon tasks with large, imperfect tool environments.</p> - https://arxiv.org/abs/2606.22388 - Sun, 21 Jun 2026 00:00:00 +0000 - - - ShotcreteDepth: A Bi-modal Dataset for Robust Robotic Depth Perception in Shotcrete Construction Environments - https://arxiv.org/abs/2606.23152 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23152.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jakub Gregorek, Lars Arnold Dethlefsen, Patrick Schmidt, Mads Essenbæk, Jonas Flink Bentzen, Lazaros Nalpantidis</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> We introduce ShotcreteDepth, a bi-modal dataset from the construction domain that captures both an active shotcreting process and general construction environments. The dataset comprises stereo RGB imagery and LiDAR point clouds acquired under harsh real-world conditions, including high turbidity and poor illumination. Such conditions adversely affect sensor measurements, leading to incomplete and noisy observations that pose significant challenges for perception systems in autonomous applications. Alongside the dataset, we release a lightweight annotation tool designed for time-efficient labeling of LiDAR point clouds. ShotcreteDepth consists of 11,252 temporally synchronized data samples, of which 220 are annotated for evaluation purposes. The dataset supports research in stereo matching, depth completion, and depth estimation under conditions that closely reflect the operational complexities found in industrial settings. Project repository: https://github.com/dtu-pas/shotcrete-depth</p> - https://arxiv.org/abs/2606.23152 + ChartWalker: Benchmarking the Cross-Chart RAG Task + https://arxiv.org/abs/2606.23997 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23997.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ning Tang, Chenghan Xie, Hanyang Yuan, Yi Li, Renhong Huang, Qian Kou, Xiaofeng Shi, Hua Zhou, Jiarong Xu</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Cross-Chart Retrieval-Augmented Generation (RAG) is critical for complex multi-modal analytical tasks in scientific, business, and political domains. However, existing benchmarks either focus on tables, which are well-structured and textualized, or generate cross-chart questions by simply extracting key points, which often induces lexical overlap between queries and evidence and yields logically inconsistent reasoning chains. To address this, we introduce ChartWalker, a novel framework for constructing challenging cross-chart RAG tasks. ChartWalker features a hierarchical knowledge graph construction method tailored to charts, which organizes entities and relations by granularity to preserve analytical structure. We then propose a structure-aware sampling algorithm that synthesizes semantically coherent, multi-hop reasoning paths, enabling explicit control over query difficulty and granularity for QA generation. Built with this framework, we release ChartWalker-Bench, a comprehensive benchmark spanning diverse domains and cross-chart query types. Extensive evaluations across major RAG paradigms reveal significant performance gaps, underscoring the benchmark's difficulty and utility. Furthermore, we provide ChartWalker-Agent, an agentic baseline to facilitate analysis and inspire future system design.</p> + https://arxiv.org/abs/2606.23997 Mon, 22 Jun 2026 00:00:00 +0000 - Lift4D: Harmonizing Single-View 3D Estimation for 4D Reconstruction In-the-Wild - https://arxiv.org/abs/2606.23688 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23688.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yehonathan Litman, Xiaoxuan Ma, Manan Shah, Nicolas Ugrinovic, Kris Kitani, Fernando De la Torre, Shubham Tulsiani</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Reconstructing dynamic non-rigid objects from monocular video requires integrating visual cues from direct observations with data-driven priors over geometry and appearance. Prior approaches either learn to directly predict 4D representations from visual input or initialize a 3D representation that is subsequently deformed and refined based on video evidence. However, the former are constrained by the scarcity of 4D training data, while the latter leverage priors only for the initial reconstruction and rely solely on video supervision thereafter; neither handles complex in-the-wild scenarios with large deformations and occlusions well. We present Lift4D, a test-time optimization framework that addresses both limitations. First, we adapt an existing single-view 3D reconstruction model to yield temporally consistent per-frame predictions via causal latent conditioning, providing a coherent initialization for a deformable 3D Gaussian Splatting representation. We then ``sculpt'' this representation to match the input video through an occlusion-aware optimization that faithfully recovers visible surface details while completing unobserved regions using a view-conditioned diffusion prior. We demonstrate that Lift4D clearly improves over prior 4D reconstruction methods, particularly on challenging in-the-wild sequences with severe occlusions and non-rigid motion.</p> - https://arxiv.org/abs/2606.23688 + VeriEvol: Scaling Multimodal Mathematical Reasoning via Verifiable Evol-Instruct + https://arxiv.org/abs/2606.23543 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23543.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoling Li, Kai Zheng, Jie Wu, Can Xu, Qingfeng Sun, Han Hu, Yujiu Yang</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Scaling reinforcement learning for visual mathematical reasoning requires more than generating harder questions: as data volume grows, the reward labels themselves must remain reliable. Yet existing data pipelines scale supervision while trusting the labeller, and policy-side methods assume the underlying answers are already correct. We instead treat scaling as a verifiable data-construction problem and decouple two axes before any policy update: prompt difficulty, expanded by route-specific evolution operators, and answer reliability, enforced by offline hypothesis-test falsification. We instantiate this as VeriEvol, an iterative framework with two extensible components: a type-aware evolution module that rewrites low-difficulty image-question seeds into harder, image-grounded prompts; and HTV-Agent, a verifier that accepts an answer only after multi-source counter-evidence has failed to refute it. The resulting verified data scales in volume, extends by adding evolution routes or verifier channels, and plugs directly into existing GRPO-style RL recipes. On a five-benchmark visual-math suite, scaling evolved SFT data from 10K to 250K samples raises the mean accuracy from 35.42 to 54.73; then, with backbone, SFT initialization, and GRPO recipe held fixed, VeriEvol adds a cumulative +3.88 over an un-evolved RL baseline, of which +1.82 comes from evolved prompts and +2.06 from the HTV-Agent verifier. We release the prompts, data, models, code, and the full verifier trace of every sample, so that downstream work can scale and audit the pipeline rather than only inspect its outputs.</p> + https://arxiv.org/abs/2606.23543 Mon, 22 Jun 2026 00:00:00 +0000 - HAKARI-Bench: A Lightweight Benchmark for Comparing Retrieval Architectures and Efficiency Settings under Unified Conditions - https://arxiv.org/abs/2606.22778 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22778.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuichi Tateno</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> With the rapid spread of retrieval-augmented generation and semantic search, choosing the right embedding and retrieval configuration is increasingly hard. Large retrieval benchmarks are comprehensive but too heavy to rerun during development, and there is little infrastructure for comparing production settings--dimensionality reduction, quantization, reranking--across many models under identical conditions. We present HAKARI-Bench, a lightweight benchmark that reconstructs existing retrieval suites into small datasets (Nano-sets): 35 benchmarks and 551 tasks across 43 languages in a unified format, enabling same-condition, model-agnostic comparison of five retrieval families (BM25, dense, sparse, late interaction, rerankers) and their efficiency variants. Across 55 models, its overall ranking reproduces the official MTEB retrieval v2, MMTEB v2 retrieval, and English BEIR (full) at Spearman >0.97. HAKARI-Bench does not replace full evaluation; it enables rapid model selection, regression detection, and reading the quality-efficiency Pareto frontier. Code, data, and leaderboard are released under the MIT license.</p> - https://arxiv.org/abs/2606.22778 + FedOT: Ownership Verification and Leakage Tracing via Watermarks for Federated LDMs + https://arxiv.org/abs/2606.22875 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22875.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wenlong Cheng, Yuan Gan, Yunqiu Xu, Jiaxu Miao</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Training Latent Diffusion Models (LDMs) within Federated Learning (FL) has attracted increasing attention due to its ability to combine the powerful generative capacity of LDMs with the privacy-preserving properties of FL. However, FL requires sharing the global model with multiple participants, which risks unauthorized model distribution or resale by malicious clients. While an intuitive approach is to adopt existing VAE-based watermarking techniques for LDMs in FL, this strategy falls short in addressing such threats due to two fundamental challenges: (1) Existing methods support ownership verification but lack the ability to trace model leakage to a specific malicious client; (2) VAE-based watermarks are vulnerable, as they can be removed simply by replacing the decoder with a clean counterpart. In this paper, we propose FedOT, the first framework for ownership verification and leakage tracing in federated LDMs. Specifically, to address the first challenge, we design a chunked watermark, where the first part is for ownership verification, and the second part is used for client identification. Furthermore, to overcome the second challenge and secure the model against VAE replacement attack, we introduce Latent Vector Transformation (LVT), which strengthens the connection between the VAE and U-Net latent spaces by modifying the original latent distribution of the VAE. Consequently, any attempt to replace the VAE for watermark removal leads to significant image quality degradation, making the LDM model unusable. Extensive experiments demonstrate that FedOT achieves superior performance in both ownership verification and traceability. Project page: https://spyzixuan.github.io/FedOT/.</p> + https://arxiv.org/abs/2606.22875 Mon, 22 Jun 2026 00:00:00 +0000 - MeshFlow: Mesh Generation with Equivariant Flow Matching - https://arxiv.org/abs/2606.23489 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23489.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qi Sun, Kiyohiro Nakayama, Jing Nathan Yan, Qixing Huang, Alexander Rush, Leonidas Guibas, Gordon Wetzstein, Jing Liao, Guandao Yang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Meshes are among the most common 3D scene representations, but directly generating meshes is challenging because the representation contains important symmetries, including permutation invariance of faces and vertices. MeshFlow learns to generate triangle meshes directly as triangle soups, avoiding the need to serialize meshes into long autoregressive sequences. We adopt equivariant optimal-transport flow matching models that respect the key symmetries of triangle soups: arbitrary permutations of faces and permutations of the vertices within each face. Toward this goal, we propose a simple yet effective modification to the Diffusion Transformer architecture, resulting in a scalable network capable of modeling a velocity field while maintaining the desired equivariance. We further introduce an optimal-transport-based training objective that improves convergence by eliminating supervision signals that violate these symmetries. MeshFlow achieves mesh quality comparable to state-of-the-art autoregressive mesh generators while providing about an 18times speedup during inference. Project page is at https://qiisun.github.io/MeshFlow/.</p> - https://arxiv.org/abs/2606.23489 + Critique of Agent Model + https://arxiv.org/abs/2606.23991 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23991.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Eric Xing, Mingkai Deng, Jinyu Hou</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> What is an agent? What constitutes agency? With the rise of Large Language Model (LLM) systems marketed as ``coding agents'', ``AI co-scientists'', and other ``agentic" tools that promise to drive up productivity, and at the same time, ``existential" concerns such as AI escaping human control with destructive power under a speculative ``machine agency" against humans, it has become essential to clarify where automation ends and agency begins, both for building capable systems and for understanding whether and what to fear. Drawing on Descartes' grounding of agency in independent thought, and on portrayals of autonomous beings in science fiction, we survey the current landscape of AI agents, and analyze agent architectures along five dimensions: goal, identity, decision-making, self-regulation, and learning. Specifically, we argue that genuine agency requires these structures to be internalized within the system itself rather than assembled through external scaffolding. This distinction between agentic systems, whose competence resides in engineered workflows, and agentive systems, whose capabilities (including social interaction) arise endogenously, defines the boundary between systems designed for prescribed tasks, and those capable of operating in the open world with true autonomy. Building on this analysis, we propose the Goal-Identity-Configurator (GIC) architecture for a general-purpose agent model, combining hierarchical goal decomposition, identity evolution, simulative reasoning grounded in a separately trained world model, learned self-regulation, and self-directed learning from both real and simulated experience. Furthermore, we share insight on the auditability, controllability, and safety of agentive systems that possess greater autonomy and ``agency", but remain under human oversight.</p> + https://arxiv.org/abs/2606.23991 Mon, 22 Jun 2026 00:00:00 +0000 - Arbor: Explicit Geometric Conditioning for Controllable 3D Asset Generation - https://arxiv.org/abs/2606.23514 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23514.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jan-Niklas Dihlmann, Andreas Engelhardt, Simon Donne, Hendrik P. A. Lensch, Mark Boss</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Text and image conditioned 3D models now generate convincing assets, but they still offer little direct control over the space an object should occupy or avoid. In authoring, this spatial intent is often known before generation starts. A chair should fit a seating envelope, a prop should leave clearance for motion, or a part should expose a contact surface. Prompts and image views are poor carriers for such constraints, requiring the need for an explicit control interface. We present Arbor, a trainable attachment for text conditioned latent 3D generation. Arbor introduces constraint meshes as a native 3D control interface. The interface uses hull regions where geometry should exist, avoidance regions that should remain empty, and touch regions the object should contact. Unlike completion or whole object scaffold control, these meshes are not target evidence. They are local typed requirements and can include regions where no surface should appear. Arbor keeps this signal as geometry by converting constraint meshes into tokens and learning a routed attachment inside a frozen denoiser. Each latent region can therefore receive the part of the constraint that matters for its spatial location. We evaluate Arbor on automatic and artist curated control benchmarks with hull, avoidance, and touch constraints, and compare the metric trends to a user preference study. Even without dedicated compliance losses, Arbor improves constraint obedience while preserving object quality and variation under fixed constraints.</p> - https://arxiv.org/abs/2606.23514 + Semantic Browsing: Controllable Diversity for Image Generation + https://arxiv.org/abs/2606.23679 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23679.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sara Dorfman, Maya Vishnevsky, Omer Dahary, Or Patashnik, Daniel Cohen-Or</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Modern text-to-image models excel in visual fidelity and prompt adherence. However, this strict adherence comes at the cost of diversity: generated samples tend to collapse into a single visual interpretation. Existing methods to improve diversity produce outputs driven by incidental variations rather than meaningful design choices. This motivates a new variant of the diversity task where structure is enforced on the generated samples. We introduce a method for controlled diversity that enables Semantic Browsing, where users can navigate structured image galleries and experience creative exploration through a systematic traversal of meaningful, interpretable axes of variation. Achieving this level of semantic control requires a deep understanding of the scene. We exploit the fact that recent text-to-image models are trained on elaborated captions, effectively decoupling semantic decision-making from pixel generation. This enables a paradigm shift: instead of relying on stochastic variation within the text-to-image model, we induce diversity directly at the text level. By leveraging rich textual representations, we allow a Vision Language Model (VLM) to operate on the full scene context. To overcome the generic outputs typical of standard VLMs, we employ an agentic workflow that explicitly enforces structured variation attuned to the original prompt. We demonstrate that our method produces diverse and navigable design spaces where every variation corresponds to a specific, user-understandable semantic decision.</p> + https://arxiv.org/abs/2606.23679 Mon, 22 Jun 2026 00:00:00 +0000 - Capable but Careless: Do Computer-Use Agents Follow Contextual Integrity? - https://arxiv.org/abs/2606.23189 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23189.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Anmol Goel, Iryna Gurevych</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Computer-use agents (CUAs) now act on a user's behalf across personal applications such as email, calendars, and to-do lists. This cross-application access is useful, but it also creates a privacy risk that has been largely overlooked: when an agent works in one context, it can pull in information from another that is inappropriate in that context. Hence, we introduce AgentCIBench, an evaluation harness that turns this risk into executable, deterministically scored scenarios. We target three common failure modes in CUAs: visual co-location, where the agent pulls in prohibited items that sit next to the task target in the UI; task-ambiguity overshare, where the agent dumps dense personal state in response to an under-specified prompt; and recipient misalignment, where the agent sends content to an addressee for whom it is inappropriate. We evaluate 15 frontier agents and find a surprisingly high failure rate: 11 of 15 leak on more than 50% of scenarios, with an average leakage of 67.9%, and the same failures persist when agents act end-to-end in the environment to complete the task. We release AgentCIBench to encourage the development of safer computer-use agents and position contextual disclosure testing as a pre-deployment safety check.</p> - https://arxiv.org/abs/2606.23189 + AOHP: An Open-Source OS-Level Agent Harness for Personalized, Efficient and Secure Interaction + https://arxiv.org/abs/2606.23449 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23449.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shanhui Zhao, Jiacheng Liu, Guohong Liu, Jichao Yan, Jialei Ye, Yuhao Yang, Hao Wen, Shizuo Tian, Yizhen Yuan, Yuxuan Chen, Yunxin Liu, Ju Ren, Ya-Qin Zhang, Chao Huang, Yao Guo, Yuanchun Li</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> AI agents are driving a new software paradigm, with the ability to autonomously call tools, extract information, manage memory, and complete tasks that span applications and data sources. Most existing end-user operating systems, however, are designed for application-centric workflows and offer little native support for AI agents. This mismatch limits the wider adoption of agents and leads to execution overhead and safety risks when running agents on conventional systems. While the concept of agent-native operating systems is emerging, the research community lacks an open testbed to explore the architectural primitives desired for agent-mediated interaction. We present AOHP (Android Open Harness Project), an OS-level agent harness built on the Android Open Source Project (AOSP). The core design principle of AOHP is to treat agents as first-class OS actors, enabling adaptive user interfaces and agent-friendly runtime environments. AOHP preserves the mature Android software and hardware ecosystem while introducing three agent-oriented system mechanisms: personalized service composition, efficient agent interfaces, and secure information flow. Based on preliminary experiments on challenging tasks covering key capabilities of OS agents, AOHP shows clear advantages in task completion (+21.12% completion rate), execution cost (-51.55% token cost), and security-policy compliance.</p> + https://arxiv.org/abs/2606.23449 Mon, 22 Jun 2026 00:00:00 +0000 - When Agents Commit Too Soon: Diagnosing Premature Commitment in LLM Agents - https://arxiv.org/abs/2606.22936 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22936.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Aman Mehta</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Long-horizon LLM agents can fail quietly: they settle on one reading of the evidence early, then spend the rest of the run defending it. We call this premature commitment. Final-answer scoring misses the failure mode because it sees only the answer, not whether the process has already collapsed to a stable path. We define representational commitment as cross-run hidden-state convergence at a fixed reasoning step, and use it as an early diagnostic of trajectory consistency. On Llama-3.1-70B running ReAct on HotpotQA, step-4 hidden-state similarity predicts downstream behavioral consistency (r = -0.35, partial r = -0.45), with a localized temporal and layer-wise signature. The signal replicates across Qwen-2.5-72B and Phi-3-14B, and on StrategyQA (r = -0.83). It does not track correctness: committed-wrong and committed-correct questions are not separable in activation similarity. That boundary is central to the claim. Commitment tells us whether an agent has settled, not whether it is right. A runtime monitor detects inconsistent trajectories from hidden states at AUROC up to 0.97 (0.85--0.88 under a stricter split), and a prompting intervention cuts behavioral variance by 28% against a token-matched control while leaving accuracy statistically unchanged. We also test whether the signal can route self-consistency compute; on a harder benchmark it helps only modestly and is matched by a simpler output-based baseline. The result is a diagnostic for a hidden process failure, with clear limits rather than a general accuracy lever.</p> - https://arxiv.org/abs/2606.22936 - Mon, 22 Jun 2026 00:00:00 +0000 + FLUX3D: High-Fidelity 3D Gaussian Generation with Diffusion-Aligned Sparse Representation + https://arxiv.org/abs/2606.24874 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24874.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haorui Ji, Weizhe Liu, Hongdong Li, Hengkai Guo</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Sparse voxel representation has emerged as a scalable foundation for image-to-3D Gaussian Splatting (3DGS) generation, yet current methods struggle to preserve high-frequency visual details of input images due to two structural bottlenecks. First, they adopt discriminative 2D features optimized for semantic abstraction to construct sparse voxel latents, which suppress reconstructive cues and induce a representation bottleneck. Second, in the generation stage, standard diffusion transformers lack effective mechanisms to align dense 2D image tokens with sparse 3D voxel latents, resulting in a cross-modal correspondence bottleneck. To address these issues, we propose FLUX3D, a scalable image-to-3DGS framework that boosts both representation learning and cross-modal alignment during generation. We first revisit 2D feature selection for sparse-voxel-based 3D representation learning, propose Diffusion-Aligned Structured Latents (DA-SLAT) and couple it with a decoder-only architecture to improve 3DGS reconstruction fidelity. We also design a sparse-structure-aware diffusion framework, which integrates the Sparse-structure Multimodal Diffusion Transformer (SMDiT) and Modal-Aware Rotary Positional Embedding (MARoPE) to achieve geometry-agnostic 2D-3D alignment. Extensive benchmark experiments demonstrate that FLUX3D yields substantial improvements in appearance fidelity and significantly outperforms all state-of-the-art (SOTA) methods in generating high-quality 3DGS assets.</p> + https://arxiv.org/abs/2606.24874 + Tue, 23 Jun 2026 00:00:00 +0000 - TROPT: An Open Framework for Unifying and Advancing Discrete Text Optimization - https://arxiv.org/abs/2606.23496 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23496.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Matan Ben-Tov, Mahmood Sharif</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Discrete text-trigger optimization -- searching for text sequences that, when ingested by a model, steer it toward a specified objective -- underpins model red-teaming (e.g., LLM jailbreaks), as well as auditing and interpretability. However, the current state of discrete optimizers hinders their adoption and progress. First, existing optimizers, when open-sourced at all, are scattered across research codebases tied to specific models, objectives, and problem domains. Second, optimizer variants proliferate, each requiring engineering overhead to use or extend, and remaining hard to compare head-to-head. Together, these raise the bar for adopting optimizers in existing or new domains, and for advancing them via new strategies. We address these gaps with TROPT, the first open-source framework that unifies discrete optimizers' execution and standardizes their development under a single interface. TROPT makes it easy to customize end-to-end optimization recipes by swapping any component -- models, objectives, and optimizers -- extending its reach across domains and new applications. TROPT currently ships with 30+ optimization recipes -- covering applications such as jailbreaking and probing model internals -- built from 15+ optimizers (spanning white-box to black-box access) and 15+ losses, from foundational to state-of-the-art methods. Demonstrating its utility, we leverage TROPT in several studies: (i) controlled, large-scale experiments comparing and enhancing optimization strategies for LLM jailbreaks, revealing potent-yet-underadopted techniques; and (ii) porting optimizers from one domain (e.g., LLM jailbreak) to new domains (e.g., corpus-poisoning embedding model). In all, TROPT significantly lowers the barrier to adopting and advancing discrete text optimization.</p> - https://arxiv.org/abs/2606.23496 - Mon, 22 Jun 2026 00:00:00 +0000 + AGORA: An Archive-Grounded Benchmark for Agentic Workplace Document Reasoning + https://arxiv.org/abs/2606.24526 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24526.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Honglin Guo, Qi Zhang, Yu Zhang, Weijie Li, Rui Zheng, Zhikai Lei, Qiyuan Peng, Zhiheng Xi, Tao Gui, Qi Zhang</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Large language models are increasingly deployed as agents that reason over documents rather than answer from parametric knowledge. We study archive-grounded reasoning: locating sparse evidence across a large, messy collection of workplace files, reconciling inconsistent terminology, units, and time conventions, and computing an answer. Existing benchmarks address only parts of this setting and none jointly stresses archive-groundedness, agentic exploration, and cross-domain coverage. We introduce Agora, a benchmark pairing 362 questions with eight domain collections of 9,664 authentic documents and 372M tokens, far exceeding any model's context window, so agents must explore deliberately rather than scan exhaustively. Agora is built by an agentic pipeline combining cross-document task synthesis, leakage-preventing obfuscation, and difficulty filtering. Evaluating eight models, we find the task far from solved: even the strongest reaches only 59.4% accuracy, with notable variation across domains.</p> + https://arxiv.org/abs/2606.24526 + Tue, 23 Jun 2026 00:00:00 +0000 - UniverSat: Resolution- and Modality-Agnostic Transformers for Earth Observation - https://arxiv.org/abs/2606.23503 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23503.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yohann Perron, Guillaume Astruc, Nicolas Gonthier, Clement Mallet, Loic Landrieu</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Vision Transformers (ViT) dominate computer vision. However, their reliance on rigid patch projectors hinders transfer to Earth Observation (EO), where input modalities, scales, and resolutions vary widely. We introduce UniverSat, a ViT-style backbone built around a Universal Patch Encoder that maps patches from arbitrary spatial, spectral, and temporal resolutions, and from both optical and non-optical sensors, into a shared embedding space with a shared set of weights. This enables training a single model on heterogeneous multimodal corpora via self-supervision, yielding robust, sensor-agnostic spatial features. We validate this approach with strong results across classification and segmentation on standard EO benchmarks from GeoBench, PANGEABench, and SpectralEarth. Our code and models are available at https://github.com/gastruc/UniverSat.</p> - https://arxiv.org/abs/2606.23503 - Mon, 22 Jun 2026 00:00:00 +0000 + InSight: Self-Guided Skill Acquisition via Steerable VLAs + https://arxiv.org/abs/2606.24884 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24884.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Maggie Wang, Lars Osterberg, Stephen Tian, Ola Shorinwa, Jiajun Wu, Mac Schwager</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Vision-language-action (VLA) models can learn manipulation skills from demonstrations, but their capabilities are bounded by the skills in the training data. We present InSight, a framework that unlocks autonomous skill acquisition by rendering VLAs steerable at the primitive-action level (e.g., "move gripper to the bowl", "lift upward", "pour the bottle"). InSight consists of two primary stages: (1) an automated segmentation pipeline that partitions demonstrations into labeled primitives via VLM plan decomposition and end-effector poses to enable VLA primitive steerability, and (2) a VLM-guided data flywheel that identifies missing primitives required to accomplish a novel task, autonomously attempts demonstrations of the missing primitives with VLM-proposed low-level control, and automatically labels, stores, and integrates successful demonstrations into the VLA training set. We evaluate InSight across simulation and real-world manipulation tasks, including block flipping, drawer closing, sweeping, twisting, and pouring, without any human demonstrations of these target skills. Once learned, these primitives can be composed to execute novel, long-horizon tasks without additional human demonstrations. Our findings demonstrate that primitive steerability provides a practical foundation for continual skill acquisition in VLA policies. Project website: https://insight-vla.github.io.</p> + https://arxiv.org/abs/2606.24884 + Tue, 23 Jun 2026 00:00:00 +0000 - Vera: A Layered Diffusion Model for Content-Preserving Video Editing - https://arxiv.org/abs/2606.23610 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23610.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hongkai Zheng, Ta-Ying Cheng, Benjamin Klein, Yisong Yue, Zhuoning Yuan</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Video diffusion models have enabled remarkable progress in video generation and editing. However, content preservation remains a core challenge: existing methods regenerate every pixel and often alter elements that should remain unchanged, such as characters or background scenes. We introduce Vera, a layered diffusion framework for content-preserving video editing. Instead of regenerating the entire video, Vera generates an edit layer along with an alpha matte for compositing with the source video, separating creative editing from content preservation by design. To encourage coherent composition with the source video, we extend the text-to-video DiT into a Mixture-of-Transformers (MoT) architecture, with separate DiTs for each layer that interact through joint self-attention. To support the training of Vera, we further construct a high-quality layered dataset with accurate alpha mattes, diverse scenes and dynamics, and visual effects. Across our quantitative benchmark and human preference study, Vera outperforms leading open-source video editing models in content preservation while remaining competitive in edit quality, using 486K frames of layered training data.</p> - https://arxiv.org/abs/2606.23610 - Mon, 22 Jun 2026 00:00:00 +0000 + World Value Models for Robotic Manipulation + https://arxiv.org/abs/2606.24742 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24742.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhihao Wang, Jianxiong Li, Yu Cui, Yuan Gao, Xianyuan Zhan, Junzhi Yu, Xiao Ma</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Generalist value models play a pivotal role in scaling robotic policy learning from large-scale, mixed-quality data. Mathematically, accurate value estimation demands deep temporal understanding, requiring models to both ground the current belief using historical context and plan over future outcomes. However, most existing robotic value models are built on Vision-Language Model (VLM) backbones that are pretrained primarily on static or temporally sparse visual observations, lacking the requisite temporal modeling capabilities for value estimation. Unlike VLMs, world models naturally excel at temporal modeling and future planning, making them ideal foundations for learning generalizable value functions. Driven by this insight, we marry world models with value estimation to construct a new generalist robotic value model, World Value Model (WVM), that offers accurate task progressions to assess data quality. On standard benchmarks, WVM delivers state-of-the-art (SOTA) Value-Order Correlation (VOC) results. Complementing standard evaluation suites that contains only expert data, we further introduce Suboptimal-Value-Bench, a multi-embodiment benchmark consisting of 800 suboptimal trajectories with high-fidelity, human-labeled frame annotations. Our evaluations show that WVM maintains its SOTA performance on Suboptimal-Value-Bench, establishing its robustness in handling both expert and suboptimal data. When deployed for policy learning, WVM improves manipulation performance across various policy extraction approaches in both simulated and real-world deployment, providing robust guidance for learning from mixed-quality data.</p> + https://arxiv.org/abs/2606.24742 + Tue, 23 Jun 2026 00:00:00 +0000 - Dense Reward for Multi-View 3D Reasoning with Global Maps and Local Views - https://arxiv.org/abs/2606.23557 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23557.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiho Choi, Seonho Lee, Seojeong Park, Hyunjung Shim</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Multi-view 3D Visual Question Answering (MV3D-VQA) requires integrating partial observations into a coherent 3D scene representation and selecting informative viewpoints for multi-step spatial reasoning. However, current multimodal LLMs are typically trained with sparse, answer-level supervision, which often yields inconsistent cross-view reasoning and brittle view selection. We present DR-MV3D (Dense Reward for MV3D-VQA), a map-grounded learning framework that provides dense, verifiable rewards to supervise the reasoning process. Our approach decomposes MV3D-VQA into (i) allocentric global map construction, (ii) question-conditioned view-trajectory planning, and (iii) egocentric grounding for answer prediction. To make intermediate steps learnable without manual annotations, we introduce two rewards: a global consistency reward that aligns the predicted map with geometry-consistent pseudo targets from frozen 3D vision foundation models (e.g., VGGT + SAM3), and a local trajectory reward that supervises ordered viewpoint selection. We optimize the full pipeline with trajectory-level policy optimization (GRPO). Experiments on MindCube, VSI-Bench, and BLINK (MV) show that DR-MV3D consistently improves over strong multi-image baselines, supporting the effectiveness of process-level dense supervision for multi-view 3D reasoning.</p> - https://arxiv.org/abs/2606.23557 - Mon, 22 Jun 2026 00:00:00 +0000 + FlowR2A: Learning Reward-to-Action Distribution for Multimodal Driving Planning + https://arxiv.org/abs/2606.24231 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24231.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xirui Li, Zhe Liu, Xiaoqing Ye, Wenhua Han, Yifeng Pan, Junyu Han, Hengshuang Zhao</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Multimodal driving planning faces a long-standing tension between two paradigms: scoring-based methods benefit from dense reward supervision but are confined to a fixed action vocabulary, while anchor-based methods generate proposals dynamically yet suffer from sparse supervision constrained to a single ground-truth trajectory. In this work, we propose FlowR2A, which resolves this tension by reframing simulation-based rewards from discriminative targets into generative conditions. By learning the reward-conditioned action distribution from dense trajectory-reward pairs with a flow-matching decoder, FlowR2A unifies the dense supervision of scoring-based methods with the proposal generation of anchor-based methods in a single generative model, forcing the model to internalize the correlation between an action and its outcomes in safety, progress, comfort, and rule compliance. To balance hard safety constraints against soft progress objectives, we introduce fine-grained per-timestep reward conditioning and reward noise augmentation. The generative formulation naturally supports controllable test-time sampling via reward guidance and anchored sampling, producing high-quality proposals. FlowR2A achieves state-of-the-art results on the NAVSIM v1 and v2 benchmarks, with multimodal proposals of substantially higher quality than prior methods.</p> + https://arxiv.org/abs/2606.24231 + Tue, 23 Jun 2026 00:00:00 +0000 - Tapered Language Models - https://arxiv.org/abs/2606.23670 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23670.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Reza Bayat, Ali Behrouz, Aaron Courville</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Modern language models, including transformer, recurrent, and memory-based variants, share a common chassis: a stack of identical layers in which parameters are allocated uniformly across depth. This is a default inherited from the original transformer and largely unchanged since, yet a growing body of evidence suggests that layers contribute non-uniformly to the final output, with later layers refining the residual stream rather than transforming it. We ask whether parameter capacity should reflect this asymmetry. Our controlled experiment shows that, under a fixed budget, allocating more capacity to earlier layers and less to later layers improves perplexity over a uniform-width baseline, while the reverse allocation hurts. Building on this result, we introduce Tapered Language Models (TLMs), an architectural principle in which a parameter-bearing component is monotonically tapered across depth under a fixed total budget. MLPs are the natural site for this instantiation: they dominate parameter count across all modern LM families and expose width as a single, clean axis of variation. Across three model scales and four architectures (Transformer, Gated Attention, Hope-attention, and Titans), tapering MLP width via a smooth cosine schedule consistently improves perplexity and downstream benchmark performance over uniform baselines, at no additional parameter or compute cost. These findings establish depth-aware capacity allocation as a simple, architecture-agnostic axis of language model design, a free lever hidden in plain sight.</p> - https://arxiv.org/abs/2606.23670 - Mon, 22 Jun 2026 00:00:00 +0000 + MEMPROBE: Probing Long-Term Agent Memory via Hidden User-State Recovery + https://arxiv.org/abs/2606.24595 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24595.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Enze Ma, Yufan Zhou, Wei-Chieh Huang, Jie Yang, Huanhuan Ma, Zixuan Wang, Chengze Li, Chunyu Miao, Philip S. Yu, Zhen Wang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Long-term memory promises LLM agents that grow more capable across sessions, maintaining an accurate, evolving understanding of the user that interaction forms. In practice, however, this memory is evaluated mostly through downstream behavior, such as later answers, personalization quality, or task success, which tests that understanding only indirectly and leaves the memory artifact itself largely unaudited. We argue that long-term memory should instead be evaluated as an auditable post-interaction artifact: after ordinary assistance, what structured user state can be reconstructed from the memory the agent leaves behind? We instantiate this view in MEMPROBE, a benchmark in which a memory-equipped agent assists simulated users, each carrying a hidden, taxonomy-anchored user-state bank, across a trajectory of leak-controlled tasks, after which that bank is reconstructed from the agent's resulting memory under both full-store and top-k access. Built on synthetic ground truth for efficient, scalable measurement, MEMPROBE spans 50 simulated users with 31 hidden dimensions each (1,550 recovery targets) and tests 5 representative memory systems. Testing state-of-the-art memory agents, we find that successful assistance and recoverable memory behave as distinct capabilities. Task completion nearly saturates, even for a memoryless baseline, while category-balanced recovery stays moderate (about 0.6) and drops further under top-k retrieval. MEMPROBE is the first benchmark to study memory recovery directly, reconstructing the user state a system retains and scoring it against ground truth. We see recovery as a concrete objective for future memory agents to optimize, and MEMPROBE as a step toward an environment where agents are trained to remember their users, growing more faithful the longer they know them.</p> + https://arxiv.org/abs/2606.24595 + Tue, 23 Jun 2026 00:00:00 +0000 - Causal Discovery in the Era of Agents - https://arxiv.org/abs/2606.23608 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23608.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yujia Zheng, Vishal Verma, Mantej Gill, Haoyue Dai, Peter Spirtes, Kun Zhang</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Recent attempts to combine large language models (LLMs) with causal discovery ask models to infer pairwise directions, propose graph structures, or inject language-model outputs as priors and constraints. These approaches promise faster analysis, but they also obscure whether a causal evidence is supported by data and assumptions or by textual associations, prompt artifacts and hallucinated mechanisms. We argue for a different role for agents in causal discovery. Agents should inspect data, retrieve context, explain method assumptions and clarify graph outputs, but they should not supply edges, orientations, priors, constraints or causal conclusions. We propose the principle that agents assist the workflow, while causal claims remain grounded in data, explicit assumptions, formal algorithms, diagnostics and user or domain-expert decisions. We instantiate this principle in causal-learn+, an online platform that coordinates data analysis, preprocessing, method recommendation, expert-knowledge incorporation, formal discovery and interpretation around the algorithmic ecosystem of causal-learn. A case study on Big Five personality data illustrates agent-assisted pipeline of causal discovery without turning language-model unreliability into causal evidence. The platform is available at causallearn.com.</p> - https://arxiv.org/abs/2606.23608 - Mon, 22 Jun 2026 00:00:00 +0000 + ReMMD: Realistic Multilingual Multi-Image Agentic Verification for Multimodal Misinformation Detection + https://arxiv.org/abs/2606.24112 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24112.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenhao Dang, Dantong Zhu, Jun Yang, Conghui He, Weijia Li</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Multimodal misinformation detection is increasingly important because viral posts now combine long multilingual narratives, several images, mixed provenance, and subtle text--image framing errors. Existing benchmarks and methods remain poorly matched to this setting: they usually isolate short captions, single images, binary labels, or one manipulation source, while agentic verification remains costly under realistic evidence search. We present ReMMD, a realistic multilingual multi-image agentic verification framework for multimodal misinformation detection. ReMMD includes ReMMDBench, a real-world multimodal misinformation detection benchmark with 500 samples, 2,756 images, five monolingual languages, two cross-lingual settings, three text-length tiers, multi-image posts, five-way veracity labels, eight distortion labels, evidence provenance, and rationales. It also includes ReMMD-Agent, a persistent-memory verifier that decomposes posts into atomic points, builds a reusable evidence set, and predicts structured L1/L2/L3 outputs. Across proprietary systems, open LVLMs, MMD-Agent, and T2-Agent, ReMMD-Agent obtains the best five-way veracity performance, with 41.80% accuracy and 39.12% macro-F1 using GPT-5.2, while reducing cost by 17.5% relative to MMD-Agent and 79.9% relative to T2-Agent. The project is available at https://dang-ai.github.io/ReMMD.</p> + https://arxiv.org/abs/2606.24112 + Tue, 23 Jun 2026 00:00:00 +0000 - Safe Few-Step Generation via Velocity Editing - https://arxiv.org/abs/2606.23267 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23267.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yujin Choi, Jaehong Yoon</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Flow matching has recently emerged as a strong paradigm for state-of-the-art text-to-image (T2I) generation, enabling high-quality generation with a small number of sampling steps. As these models are increasingly integrated into real-world applications, ensuring safe and non-sensitive content generation has become a critical requirement. However, adapting safety and concept removal methods to this new generation framework remains an open challenge. Specifically, prior methods largely rely on iterative trajectory steering across a number of denoising steps or on CLIP-centric prompt embedding manipulation. These design assumptions pose fundamental bottlenecks for safety in flow matching-based T2I generation, where limited sampling steps constrain iterative correction and modern context-aware text encoders diminish the effectiveness of embedding-level interventions. In this paper, we propose VESFlow, a training-free safety method tailored to flow matching with extremely few sampling steps. Leveraging the fact that flow matching models learn the marginal velocity, we directly edit the velocity field via a safe-conditional posterior. VESFlow steers the trajectory toward safe outputs while leaving the conditioning prompt unchanged. Building on the observation that VESFlow leaves outputs unchanged under benign prompts, we further introduce a risk score-based filtering that bypasses velocity editing to reduce computational cost while preserving benign prompt generation. Based on this filtering, we propose VESFlow+, a stronger variant of VESFlow that not only edits the velocity toward the safe direction, but also pushes it away from the unsafe direction. Experimental results show that VESFlow+ removes the target concept, reducing the attack success rate by NudeNet to 6.3% on Ring-A-Bell and 6.8% on MMA-Diffusion on the 4-step MeanFlow model, while preserving fidelity on benign prompts.</p> - https://arxiv.org/abs/2606.23267 - Mon, 22 Jun 2026 00:00:00 +0000 + DREAM: Dense Retrieval Embeddings via Autoregressive Modeling + https://arxiv.org/abs/2606.24667 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24667.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yixuan Tang, Yi Yang</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Dense retrieval embedding models are a fundamental component of modern retrieval-based AI systems. Most dense retrievers are trained with contrastive objectives, which require labeled positive and negative document pairs that are often costly and difficult to obtain. In this work, we investigate whether the autoregressive next-token prediction objective of a large language model (LLM) can provide supervision for dense retrieval. The intuition is simple: if a document contains information relevant to a query, conditioning on that document should make the target output easier for the LLM to predict. A key challenge is that the next-token prediction loss is computed inside the LLM, while the retriever is a separate embedding model. To address this challenge, we propose DREAM (Dense Retrieval Embeddings via Autoregressive Modeling), which injects retriever-generated query-document similarity scores into selected attention heads of a frozen LLM. During training, these scores determine how much attention each candidate document receives while the LLM predicts the target output. The resulting prediction loss provides gradients for retriever training through the attention mechanism. We evaluate DREAM on retrieval benchmarks BEIR and RTEB using embedding backbones ranging from 0.5B to 3B parameters. DREAM consistently outperforms existing baselines across different model scales. These results demonstrate that DREAM provides a promising approach for training dense retrievers through autoregressive modeling.</p> + https://arxiv.org/abs/2606.24667 + Tue, 23 Jun 2026 00:00:00 +0000 - Tmax: A simple recipe for terminal agents - https://arxiv.org/abs/2606.23321 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23321.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hamish Ivison, Junjie Oscar Yin, Rulin Shao, Teng Xiao, Nathan Lambert, Hannaneh Hajishirzi</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Terminal-using agents have quickly become the most popular downstream application of language models (LMs). Despite their prevalence, relatively little academic work has examined RL-based training of these models, likely due to difficult benchmarks, a lack of data, and a lack of simple baseline recipes. We present Tmax, the strongest open RL recipe for terminal agents to date, bringing open data recipes closer to the frontier. While simple, our recipe achieves 27\% on Terminal-Bench 2.0 with only 9B parameters, outperforming much larger models from prior work. Concretely, we generate data using a novel taxonomy, combining difficulty control, personas, and verifier diversification, which allows us to cheaply generate large amounts of terminal environments for RL and SFT training. We open-source our terminal dataset, which is over 2.5x larger than previously released terminal-agent datasets. We then train open-weight models using RL with our data, using a simple, outcome-only recipe. We release our data, models, and code as a strong baseline for future open academic work on terminal agents at https://github.com/hamishivi/tmax.</p> - https://arxiv.org/abs/2606.23321 - Mon, 22 Jun 2026 00:00:00 +0000 + Holistic Data Scheduler for LLM Pre-training via Multi-Objective Reinforcement Learning + https://arxiv.org/abs/2606.24133 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24133.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenhao Dang, Jing Ma, Mingjie Liao</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> The composition of training data, governed by the diversity of sources and their mixing strategy, is a cornerstone of Large Language Model (LLM) pre-training. Online Data Mixing (ODM), the technique of adaptively adjusting data mixtures during training, has emerged as a promising direction to improve efficiency. However, existing methods are constrained by their reliance on a singular optimization perspective, which fundamentally overlooks the need for complex LLM pre-training to consider the dynamic data composition from multiple dimensions. To overcome this limitation, we introduce the Holistic Data Scheduler (HDS), a novel online data mixing framework. HDS formulates the data scheduling challenge as a reinforcement learning problem in a continuous control space and leverages the Soft Actor-Critic (SAC) algorithm for its stability and sample efficiency in exploring the high-dimensional policy space. At the core of HDS lies a novel multi-objective, holistic reward function that integrates three critical perspectives: a data-driven reward for quality, a loss-driven reward capturing inter-domain influence, and a model-driven reward based on weight norms. To validate our design and determine its optimal configuration, we conducted systematic experiments on LLMs of various sizes. On The Pile benchmark, HDS reaches the final validation perplexity of the next best method with 44% fewer training iterations. Furthermore, it achieves a 7.2% improvement on the MMLU 0-shot task along with consistent gains on other benchmarks, showcasing its ability to enhance both training efficiency and final model capability.</p> + https://arxiv.org/abs/2606.24133 + Tue, 23 Jun 2026 00:00:00 +0000 - Training Open Models for Agentic Phone Use - https://arxiv.org/abs/2606.23049 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23049.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhengyang Tang, Xin Lai, Pengyuan Lyu, Xinyuan Wang, Tianyi Bai, Chenxin Li, Yiduo Guo, Huawen Shen, Yuxuan Liu, Junyi Li, Zhengyao Fang, Yang Ding, Yi Zhang, Weinong Wang, Xingran Zhou, Liang Wu, Fei Tang, Sunqi Fan, Shangpin Peng, Zheng Ruan, Anran Zhang, Benyou Wang, Ji-Rong Wen, Rui Yan, Chengquan Zhang, Han Hu</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Phones are becoming an important execution surface for general-purpose agents, but training open models for reliable phone use remains difficult because the environment that matters at deployment, real devices running real apps, is slow, stateful, side-effectful, and hard to reset or verify, while scalable mock environments only approximate real behavior. We present PhoneBuddy, a training recipe and open-model line for agentic phone use that combines a real-app environment with a mock-app environment, PhoneWorld, which reconstructs runnable mock apps from real GUI usage structure. PhoneBuddy first builds a shared supervised fine-tuning stage from trajectories collected in both environments, then compares real-app RL against mixed RL across both environments. Across a 150-task human evaluation on real phones spanning apps, mini-apps, and cross-app workflows, task success rate improves from 36.67\% after supervised fine-tuning to 40.67\% after real-app RL and 45.33\% after mixed RL. On AndroidWorld, the same progression rises from 60.3\% to 77.2\% to 83.2\%. These results show that mock-app training is not a replacement for real-app RL, but a complementary source of scalable, resettable, and automatically checked interaction. The gains are strongest on app and mini-app tasks, while long-horizontal cross-app workflows remain an important open challenge.</p> - https://arxiv.org/abs/2606.23049 - Mon, 22 Jun 2026 00:00:00 +0000 + DiffusionBench: On Holistic Evaluation of Diffusion Transformers + https://arxiv.org/abs/2606.24888 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24888.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xingjian Leng, Jaskirat Singh, Zhanhao Liang, Ethan Smith, Martin Bell, Aninda Saha, Yuhui Yuan, Liang Zheng</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Diffusion transformer (DiT) research on image generation has converged to a single evaluation setup: class-conditional generation on ImageNet. While methods improve the FID and related metrics, it is increasingly unclear whether they reflect real progress in generative modeling. The natural alternative, i.e., text-to-image (T2I) generation, is perceived as too costly or inconvenient to train and evaluate and is often skipped. We argue that this perception no longer holds. We introduce NanoGen, a unified DiT training and evaluation framework. NanoGen matches state-of-the-art DiT baselines on ImageNet and, with 12 lines of configuration change, also trains competitive text-to-image models. It currently supports RAE, VAE, pixel-space, and MeanFlow diffusion methods under both ImageNet and T2I setups. Under NanoGen, training T2I requires comparable compute to ImageNet. After training 21 latent diffusion models with NanoGen, we observe that method ranking shows no strong correlation between ImageNet and T2I generation: Pearson correlation is between -0.377 and -0.580 across three metrics. This suggests that a method which improves class-conditional ImageNet FID may show no corresponding improvement on T2I, clearly indicating the necessity of evaluating DiTs on both tasks. To this end, we summarize ImageNet and text-to-image results, which yields DiffusionBench, a holistic benchmark for DiT research. We recommend reporting DiffusionBench in place of ImageNet alone: methods that improve DiffusionBench are more likely to reflect broader progress.</p> + https://arxiv.org/abs/2606.24888 + Tue, 23 Jun 2026 00:00:00 +0000 - Self-Compacting Language Model Agents - https://arxiv.org/abs/2606.23525 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23525.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tianjian Li, Jingyu Zhang, William Jurayj, Xi Wang, Chuanyang Jin, Mehrdad Farajtabar, Eric Nalisnick, Daniel Khashabi</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> Long agent traces composed of chains of thought and tool calls accumulate stale content that anchor subsequent generations, and eventually outgrow the context window. Existing scaffolds mitigate it with fixed-interval compaction triggered at a token threshold. Such triggers pay no heed to trajectory structure, risking discard of partial results mid-derivation or mid-search. We propose SelfCompact, a scaffold that allows the model itself to decide when and how to compact. Specifically, it pairs two inference-time elements: (i) a compaction tool the model invokes to summarize the accumulated context, and (ii) a lightweight rubric specifying when to fire (a sub-task has resolved, or the trajectory is converging) and when to suppress (mid-derivation, or when stuck). Both are needed. The tool alone is unevenly used across open-weight models, often invoked at unhelpful moments or not at all; the rubric alone cannot act. Together, they elicit effective adaptive compaction without any fine-tuning or external supervision. We present empirical results on six benchmarks (competitive math and agentic search) and seven models. Our results show that SelfCompact matches or exceeds fixed-interval summarization at a fraction of the token cost, improving over a no-summarization baseline by up to 18.1 points on math and 5-9 points on agentic search at 30-70% lower per-question cost. Our results expose a meta-cognitive gap: although unprompted models cannot reliably tell when their own context is rotting, a lightweight rubric closes this gap, reframing when to compact as a capability that scaffolds can supply without training.</p> - https://arxiv.org/abs/2606.23525 - Mon, 22 Jun 2026 00:00:00 +0000 + Are Text-to-Image Models Inductivist Turkeys? A Counterfactual Benchmark for Causal Reasoning + https://arxiv.org/abs/2606.24548 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24548.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiayi Lei, Yuandong Pu, Xingyu Han, Rongpeng Zhu, Jing Xu, Jinyao Wang, Zijian Zhou, Bin Fu, Yuewen Cao, Yihao Liu, Yongsheng Li</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Text-to-image (T2I) generation models have achieved remarkable progress in producing visually realistic images from natural language prompts. Yet it remains unclear whether their success reflects genuine causal understanding or sophisticated pattern matching over visual-textual correlations. Inspired by Russell's inductivist turkey, we introduce Counterfactual-World (CF-World), a counterfactual benchmark designed to investigate whether text-to-image models can generate images under rules that systematically contradict real-world priors. CF-World organizes each scenario into three progressive levels: factual generation under ordinary world knowledge, explicit counterfactual generation with direct visual instructions, and implicit counterfactual generation requiring causal deduction from altered rules. We evaluate both open-source and closed-source T2I models using a Vision Language Model (VLM)-based evaluator (CF-Eval). Furthermore, we introduce two metrics: Prior Resistance Rate (PRR), which measures a model's ability to overcome entrenched real-world priors, and Reasoning Retention Rate (RRR), which assesses whether models can maintain reasoning-dependent counterfactual generation without explicit visual cues. Experiments show that all models exhibit sharp degradation from factual to counterfactual settings. Further analyses suggest that these failures arise because current T2I models encode world knowledge and visual appearances as tightly coupled patterns. Consequently, their heavy reliance on frequent visual co-occurrences within the training data forces them to default to familiar commonsense priors when tasked with rendering counterfactual worlds.</p> + https://arxiv.org/abs/2606.24548 + Tue, 23 Jun 2026 00:00:00 +0000 - Foresight: Failure Detection for Long-Horizon Robotic Manipulation with Action-Conditioned World Model Latents - https://arxiv.org/abs/2606.23085 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23085.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoran Zhang, Yifu Lu, Boyang Wang, Xuhui Kang, Yen-Ling Kuo, Zezhou Cheng, Mengdi Wang, Odest Chadwicke Jenkins</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> Long-horizon tasks are common in real-world robotic deployments, yet failure detection for such tasks remains underexplored. Detecting failures in long-horizon robotic tasks is particularly challenging because failure onset is often ambiguous and dense temporal annotations are typically unavailable. We present Foresight, a failure detection framework that monitors manipulation trajectories using latent representations from an action-conditioned world model. Foresight is trained using only final task-level success or failure labels. By leveraging predictive world-model embeddings, our method provides a unified framework for failure detection across different policies. We further use functional conformal prediction (FCP) to calibrate detection thresholds adaptively. We evaluate Foresight with state-of-the-art vision-language-action policies in simulation on LIBERO-Long, ManiSkill-Long, and BEHAVIOR-1K, compare it against state-of-the-artfailure detection methods, and validate it on real robots with three long-horizon tasks on a ReactorX-200 arm and one task on a Franka arm. Our results suggest that action-conditioned world-model embeddings provide a scalable representation for reliable failure monitoring in long-horizon manipulation.</p> - https://arxiv.org/abs/2606.23085 - Mon, 22 Jun 2026 00:00:00 +0000 + Escaping the Self-Confirmation Trap: An Execute-Distill-Verify Paradigm for Agentic Experience Learning + https://arxiv.org/abs/2606.24428 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24428.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shiding Zhu, Yudi Qi, Yajie Wang, Jiaze Li, Chao Song, Yaorui Shi, Yibo Miao, Hanqi Gao, Kai Zhang</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Experience-driven self-evolution is critical for large language model (LLM) agents to improve through open-world interaction. However, existing experience learning methods mostly rely on single-agent loops, where the same agent executes tasks, summarizes outcomes, and determines memory content. This setup makes agents vulnerable to the Self-Confirmation Trap: wrong-but-self-consistent trajectories are misidentified as successful experience, leading to cumulative errors during retrieval and reuse. To address this issue, we propose EDV, an Execute-Distill-Verify framework for reliable experience learning. In the Execute stage, multiple heterogeneous agents explore the same task space in parallel to generate diverse candidate trajectories. In the Distill stage, a dedicated third-party agent comparatively analyzes these trajectories to produce candidate experiences, reducing executor-centric summarization bias. In the Verify stage, the execution group validates candidates via a consensus mechanism, and only approved experiences are written into shared or private memory. By decoupling the three stages, EDV transforms experience learning from isolated self-reflection into collaborative construction, filtering erroneous and noisy content before memory insertion. We evaluate EDV on three challenging long-horizon benchmarks: tau2-bench, Mind2Web and MMTB. Results show EDV consistently outperforms strong baselines, validating that reliable experience construction is essential for robust agent self-evolution. Our code is available at https://github.com/shidingz/EDV.</p> + https://arxiv.org/abs/2606.24428 + Tue, 23 Jun 2026 00:00:00 +0000 - Unlimited OCR Works - https://arxiv.org/abs/2606.23050 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23050.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Youyang Yin, Huanhuan Liu, YY, Qunyi Xie, Chaorun Liu, Shiqi Yang, Shaohua Wang, Zhanlong Liu, Hao Zou, Jinyue Chen, Shu Wei, Jingjing Wu, Mingxin Huang, Zhen Wu, Guibin Wang, Tengyu Du, Lei Jia</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Recently, end-to-end OCR models, exemplified by DeepSeek OCR, have once again thrust OCR into the spotlight. A widely held view is that employing a large language model (LLM) as the decoder allows the model to leverage the prior distribution of language, leading to improved OCR performance. However, the downside is equally evident: as the output sequence lengthens, the accumulated KV cache drives up memory consumption and progressively slows down generation. This stands in stark contrast to humans, who exhibit no such decline in efficiency during long-horizon copying tasks. In this technical report, we propose Unlimited OCR, a model designed to emulate human parsing working memory. Taking DeepSeek OCR as the baseline, we replace all attention layers in the decoder with our proposed Reference Sliding Window Attention (R-SWA), which reduces attention computation costs while maintaining a constant KV cache throughout the entire decoding process. By combining the high compression rate of DeepSeek OCR's encoder with our constant KV cache design, Unlimited OCR can transcribe dozens of pages of documents in a single forward pass under a standard maximum length of 32K. More importantly, R-SWA is a general-purpose parsing attention mechanism - beyond OCR, it is equally applicable to tasks such as ASR, translation, etc. Codes and model weights are publicly available at http://github.com/baidu/Unlimited-OCR.</p> - https://arxiv.org/abs/2606.23050 - Mon, 22 Jun 2026 00:00:00 +0000 + FLAT: Feedforward Latent Triangle Splatting for Geometrically Accurate Scene Generation + https://arxiv.org/abs/2606.24876 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24876.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Orest Kupyn, Goutam Bhat, Philipp Henzler, Fabian Manhardt, Christian Rupprecht, Federico Tombari</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Generating explorable 3D scenes from a single image requires strong generative priors and accurate geometric representations suitable for downstream use. Current video diffusion models offer high-quality generation and implicitly encode multi-view geometric structure in latent space. However, existing feedforward latent scene decoders typically output volumetric 3D Gaussians that lack a well-defined surface, limiting their use in simulation or standard graphics pipelines. This motivates decoding surface-aligned primitives that are not only renderable but also closer to explicit geometric assets. We ask whether compressed video diffusion latents can be mapped directly to explicit surface primitives in a single pass. To this end, we introduce FLAT and, for the first time, show that triangle splats can be decoded directly from video diffusion latents. Compared with decoding 3D Gaussians, predicting flat primitives is notoriously more challenging due to high sensitivity to primitive orientations, oftentimes leading to poor gradient flow. FLAT solves with two key ingredients: a ray-centered rotation parameterization for triangle regression and a novel product window function that improves gradient flow during differentiable triangle rendering. On standard benchmarks, FLAT achieves significantly better geometric accuracy while maintaining competitive visual quality compared to state-of-the-art feedforward baselines. We further show that a lightweight test-time refinement step converts the predicted triangle soup into a fully opaque, game-engine-ready representation that supports real-time rendering. By evaluating 3DGS, 2DGS, and triangle splatting variants under an identical training setup, we provide the first systematic analysis of representation tradeoffs in feedforward scene generation. The project page is available at https://flat-splat.github.io</p> + https://arxiv.org/abs/2606.24876 + Tue, 23 Jun 2026 00:00:00 +0000 - CLI-Universe: Towards Verifiable Task Synthesis Engine for Terminal Agents - https://arxiv.org/abs/2606.22883 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22883.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhanbo Hua, Yifan Yao, Weihao Xie, Yongchi Zhao, Minghao Liu, Ruizhi Qiu, Zhewei Huang, Zun Wang, Yiyan Ji, Yunhai Ye, Letian Zhu, Xinping Lei, Han Li, Zhiyuan Ma, Zili Wang, Zhaoxiang Zhang, Jiaheng Liu</p><p><b>Upvotes:</b> 27</p><p><b>Summary:</b> While recent LLM-based terminal agents have demonstrated promising capabilities, the scarcity of high-quality, executable training data remains a critical bottleneck. Existing synthesis pipelines typically scale by retrofitting surface-level artifacts into tasks, frequently yielding ambiguous instructions, shallow execution paths, and brittle tests that provide weak learning signals. To overcome this, we introduce CLI-Universe, a principled synthesis engine that constructs terminal-agent tasks. CLI-Universe generates candidate tasks by sampling combinations across a multi-dimensional capability taxonomy (domain, skill type, capability, and engineering pillar), then grounds each candidate through evidence-guided deep research over real-world technical materials. To ensure rigorous supervision, validated blueprints are instantiated into Dockerized environments and subjected to a multi-stage executable verification pipeline featuring rubric-gated test construction, hint-conditional filtering, and strict fail-to-pass checking. Across the full pipeline, from candidate generation to verification, approximately two-thirds of candidates are discarded, retaining only those that are genuine, verifiable, and non-trivially challenging. To validate our framework, we instantiate a highly distilled dataset of 6,000 trajectories called CLI-Universe-6K. Remarkably, fine-tuning Qwen3-32B on CLI-Universe-6K achieves 33.4% on Terminal-Bench 2.0. This sets a new state-of-the-art for models trained on open-source data at or below 32B parameters, and outperforms several models an order of magnitude larger, demonstrating the profound data efficiency of structured, high-fidelity synthesis.</p> - https://arxiv.org/abs/2606.22883 - Mon, 22 Jun 2026 00:00:00 +0000 + OpenThoughts-Agent: Data Recipes for Agentic Models + https://arxiv.org/abs/2606.24855 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24855.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Negin Raoof, Richard Zhuang, Marianna Nezhurina, Etash Guha, Atula Tejaswi, Ryan Marten, Charlie F. Ruan, Tyler Griggs, Alexander Glenn Shaw, Hritik Bansal, E. Kelly Buchanan, Artem Gazizov, Reinhard Heckel, Chinmay Hegde, Sankalp Jajee, Daanish Khazi, Emmanouil Koukoumidis, Xiangyi Li, Hange Liu, Shlok Natarajan, Harsh Raj, Nicholas Roberts, Ethan Shen, Nishad Singhi, Michael Siu, Ashima Suvarna, Hanwen Xing, Patrick Yubeaton, Robert Zhang, Leon Liangyu Chen, Xiaokun Chen, Steven Dillmann, Saadia Gabriel, Xunyi Jiang, Anurag Kashyap, Boxuan Li, Yein Park, Minh Pham, Sujay Sanghavi, Lin Shi, Ke Sun, Yixin Wang, Zhiwei Xu, Erica Zhang, Siyan Zhao, Wanjia Zhao, Jenia Jitsev, Alex Dimakis, Benjamin Feuer, Ludwig Schmidt</p><p><b>Upvotes:</b> 27</p><p><b>Summary:</b> Agentic language models dramatically expand the applications of AI yet little is publicly known about how to curate training data for broadly capable agents. Existing open efforts such as SWE-Smith, SERA, and Nemotron-Terminal typically target a single benchmark, leaving open the question of how to train models that generalize across diverse agentic tasks. The OpenThoughts-Agent (OT-Agent) project addresses this gap with a fully open data curation pipeline for training agentic models. We conduct more than 100 controlled ablation experiments to systematically investigate each stage of the pipeline, yielding insights on the importance of task sources and diversity. We then assemble a training set of 100K examples from our pipeline and fine-tune Qwen3-32B on this dataset, which yields an average accuracy of 44.8% across seven agentic benchmarks and a 3.9 percentage point improvement over the strongest existing open data agentic model (Nemotron-Terminal-32B, 40.9%). Moreover, our training data exhibits strong scaling properties, outperforming alternative open datasets at every training set size in compute-controlled comparisons. We publicly release our training sets, data pipeline, experimental data, and models at openthoughts.ai to support future open research on agentic model training.</p> + https://arxiv.org/abs/2606.24855 + Tue, 23 Jun 2026 00:00:00 +0000 - KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking - https://arxiv.org/abs/2606.22807 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.22807.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xinping Zhao, Jiaxin Xu, Ziqi Dai, Xin Zhang, Shouzheng Huang, Danyu Tang, Xinshuo Hu, Meishan Zhang, Baotian Hu, Min Zhang</p><p><b>Upvotes:</b> 39</p><p><b>Summary:</b> As retrieval systems scale, high-quality reranking becomes increasingly important. However, most existing rerankers, whether encoder-based or decoder-based, jointly encode the query and passage, tightly coupling their computation and limiting deployment efficiency as well as flexibility. We present KaLM-Reranker-V1, a fast but not late-interaction (FBNL) reranker that decouples query and passage computation while retaining expressive relevance modeling. Built on an encoder-decoder architecture, KaLM-Reranker-V1 uses the encoder to pre-encode passages with Matryoshka embedding pooling, while the decoder models the system instruction, user instruction, and query intent; cross-attention then captures relevance between the query context and passage representations. This design makes KaLM-Reranker-V1 efficient through decoupled passage encoding, yet not late interaction, by preserving rich relevance modeling through cross-attention. We instantiate KaLM-Reranker-V1 in three sizes, Nano, Small, and Large, with 0.27B, 1B, and 4B activated parameters, respectively. Extensive experiments on BEIR, MIRACL, and LMEB demonstrate that KaLM-Reranker-V1 achieves strong reranking performance with superior efficiency. On BEIR, KaLM-Reranker-V1 achieves state-of-the-art performance, on par with strong industrial models such as the Qwen3-Reranker series; on MIRACL, despite not being extensively trained on multilingual data, KaLM-Reranker-V1 still shows excellent reranking performance. Moreover, on LMEB, reranking models demonstrate a clear advantage, with even the 0.27B Nano model remaining competitive with 7-12B embedding models.</p> - https://arxiv.org/abs/2606.22807 - Mon, 22 Jun 2026 00:00:00 +0000 + NatureBench: Can Coding Agents Match the Published SOTA of Nature-Family Papers? + https://arxiv.org/abs/2606.24530 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24530.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuru Wang, Lejun Cheng, Yuxin Zuo, Sihang Zeng, Bingxiang He, Che Jiang, Junlin Yang, Yuchong Wang, Kaikai Zhao, Weifeng Huang, Kai Tian, Zhenzhao Yuan, Jincheng Zhong, Weizhi Wang, Ning Ding, Bowen Zhou, Kaiyan Zhang</p><p><b>Upvotes:</b> 48</p><p><b>Summary:</b> We introduce NatureBench, a cross-discipline benchmark of 90 tasks distilled from peer-reviewed Nature-family publications, designed to evaluate whether AI coding agents can move beyond reproduction toward discovery on real scientific problems. NatureBench is built on NatureGym, an automated pipeline that constructs a standardized, per-task containerized environment from a source paper, addressing the environment-fragmentation problem that has limited the credibility of prior agent-on-research benchmarks. Evaluating ten frontier agent configurations under a strict web-search-disabled protocol, we find that the strongest model surpasses SOTA on only 17.8% of tasks under the g>0.1 criterion. Analysis of method pathways reveals that agents succeed primarily through methodological translation, converting scientific tasks into familiar supervised prediction problems, rather than through genuine scientific invention. Failures are dominated by wrong method choice and insufficient compute budget, not by task misunderstanding. We release the benchmark, the NatureGym pipeline, and a public leaderboard with maintainer-side reproduction. Code: https://github.com/FrontisAI/NatureBench</p> + https://arxiv.org/abs/2606.24530 + Tue, 23 Jun 2026 00:00:00 +0000 - EnterpriseClawBench: Benchmarking Agents from Real Workplace Sessions - https://arxiv.org/abs/2606.23654 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.23654.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jincheng Zhong, Weizhi Wang, Che Jiang, Kai Tian, Zhenzhao Yuan, Junlin Yang, Dianqiao Lei, Kaiyan Zhang</p><p><b>Upvotes:</b> 56</p><p><b>Summary:</b> Enterprise agents increasingly operate inside workspaces: they read heterogeneous files, invoke tools, and deliver business artifacts. We introduce EnterpriseClawBench, an enterprise agent benchmark constructed from proprietary, real-world agent sessions. Starting from a large archive of workplace sessions, the EnterpriseClawBench produces 852 reproducible tasks, each paired with recovered fixtures, rewritten prompts, role classes, skill subclasses, hard rules, and semantic rubrics. Because the sessions contain internal enterprise content, we do not release the benchmark data; instead, our reusable contribution is the construction and evaluation protocol. On EnterpriseClawBench, the best configuration reaches only 0.663 (Codex with GPT-5.5). These results show that enterprise agent evaluation must report harness--model combinations, artifact delivery, visual quality, cost, runtime, and skill-transfer behavior, rather than collapsing performance into a single score. Code: https://github.com/FrontisAI/EnterpriseClawBench</p> - https://arxiv.org/abs/2606.23654 - Mon, 22 Jun 2026 00:00:00 +0000 + Qwen-AgentWorld: Language World Models for General Agents + https://arxiv.org/abs/2606.24597 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.24597.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuxin Zuo, Zikai Xiao, Li Sheng, Fei Huang, Jianhong Tu, Yuxuan Liu, Tianyi Tang, Xiaomeng Hu, Yang Su, Qingfeng Lan, Yantao Liu, Qin Zhu, Yinger Zhang, Bowen Yu, Haiquan Zhao, Haiyang Xu, Jianxin Yang, Jiayang Cheng, Junyang Wang, Lianghao Deng, Mingfeng Xue, Tianyi Bai, Yang Fan, Yubo Ma, Yucheng Li, Zeyu Cui, Zhihai Wang, Zhihui Xie, Zhuorui Ye, An Yang, Dayiheng Liu, Jingren Zhou, Ning Ding</p><p><b>Upvotes:</b> 87</p><p><b>Summary:</b> A world model predicts environment dynamics based on current observations and actions, serving as a core cognitive mechanism for reasoning and planning. In this work, we investigate how world modeling based on language models can further push the boundaries of general agents. (i) We first focus on building foundation models for agentic environment simulation. We introduce Qwen-AgentWorld-35B-A3B and Qwen-AgentWorld-397B-A17B, the first language world models capable of simulating agentic environments covering 7 domains via long chain-of-thought reasoning. Leveraging more than 10M environment interaction trajectories of 7 domains in real-world environments, we develop Qwen-AgentWorld through a three-stage training pipeline: CPT injects general-purpose world modeling capabilities from the state transition dynamics and augmented professional corpora, SFT activates next-state-prediction reasoning, and RL sharpens simulation fidelity through a tailored framework with hybrid rubric-and-rule rewards. To evaluate language world models, we present AgentWorldBench, a comprehensive benchmark constructed from real-world interactions of 5 frontier models on 9 established benchmarks. Empirical results demonstrate that Qwen-AgentWorld significantly outperforms existing frontier models. (ii) Beyond foundation models, we further investigate two complementary paradigms through which world modeling enhances general agents. First, as a decoupled environment simulator, Qwen-AgentWorld supports scalable and controllable simulation of thousands of real-world environments for agentic RL, yielding gains that surpass real-environment training alone. Second, as a unified agent foundation model, world-model training acts as a highly effective warm-up that improves downstream performance across 7 agentic benchmarks. Code: https://github.com/QwenLM/Qwen-AgentWorld</p> + https://arxiv.org/abs/2606.24597 + Tue, 23 Jun 2026 00:00:00 +0000