diff --git a/feed_weekly.xml b/feed_weekly.xml index 05bb002..e9b0814 100644 --- a/feed_weekly.xml +++ b/feed_weekly.xml @@ -7,566 +7,741 @@ http://www.rssboard.org/rss-specification python-feedgen en - Mon, 05 Jan 2026 00:13:53 +0000 + Mon, 12 Jan 2026 00:13:35 +0000 - GRAN-TED: Generating Robust, Aligned, and Nuanced Text Embedding for Diffusion Models - https://arxiv.org/abs/2512.15560 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.15560.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bozhou Li, Sihan Yang, Yushuo Guan, Ruichuan An, Xinlong Chen, Yang Shi, Pengfei Wan, Wentao Zhang, Yuanxing zhang</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> The text encoder is a critical component of text-to-image and text-to-video diffusion models, fundamentally determining the semantic fidelity of the generated content. However, its development has been hindered by two major challenges: the lack of an efficient evaluation framework that reliably predicts downstream generation performance, and the difficulty of effectively adapting pretrained language models for visual synthesis. To address these issues, we introduce GRAN-TED, a paradigm to Generate Robust, Aligned, and Nuanced Text Embeddings for Diffusion models. Our contribution is twofold. First, we propose TED-6K, a novel text-only benchmark that enables efficient and robust assessment of an encoder's representational quality without requiring costly end-to-end model training. We demonstrate that performance on TED-6K, standardized via a lightweight, unified adapter, strongly correlates with an encoder's effectiveness in downstream generation tasks. Notably, under our experimental setup, compared with training a diffusion model from scratch, evaluating with TED-6K is about 750times faster. Second, guided by this validated framework, we develop a superior text encoder using a novel two-stage training paradigm. This process involves an initial fine-tuning stage on a Multimodal Large Language Model for better visual representation, followed by a layer-wise weighting method to extract more nuanced and potent text features. Our experiments show that the resulting GRAN-TED encoder not only achieves state-of-the-art performance on TED-6K but also leads to demonstrable performance gains in text-to-image and text-to-video generation. Our TED-6K dataset and evaluation code are available at the following link: https://anonymous.4open.science/r/GRAN-TED-4FCC/.</p> - https://arxiv.org/abs/2512.15560 - Wed, 17 Dec 2025 16:09:43 +0000 + Can LLMs Predict Their Own Failures? Self-Awareness via Internal Circuits + https://arxiv.org/abs/2512.20578 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20578.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Amirhosein Ghasemabadi, Di Niu</p><p><b>Upvotes:</b> 69</p><p><b>Summary:</b> Large language models (LLMs) generate fluent and complex outputs but often fail to recognize their own mistakes and hallucinations. Existing approaches typically rely on external judges, multi-sample consistency, or text-based self-critique, which incur additional compute or correlate weakly with true correctness. We ask: can LLMs predict their own failures by inspecting internal states during inference? We introduce Gnosis, a lightweight self-awareness mechanism that enables frozen LLMs to perform intrinsic self-verification by decoding signals from hidden states and attention patterns. Gnosis passively observes internal traces, compresses them into fixed-budget descriptors, and predicts correctness with negligible inference cost, adding only ~5M parameters and operating independently of sequence length. Across math reasoning, open-domain question answering, and academic knowledge benchmarks, and over frozen backbones ranging from 1.7B to 20B parameters, Gnosis consistently outperforms strong internal baselines and large external judges in both accuracy and calibration. Moreover, it generalizes zero-shot to partial generations, enabling early detection of failing trajectories and compute-aware control. These results show that reliable correctness cues are intrinsic to generation process and can be extracted efficiently without external supervision.</p> + https://arxiv.org/abs/2512.20578 + Tue, 23 Dec 2025 18:21:32 +0000 - Mindscape-Aware Retrieval Augmented Generation for Improved Long Context Understanding - https://arxiv.org/abs/2512.17220 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17220.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuqing Li, Jiangnan Li, Zheng Lin, Ziyan Zhou, Junjie Wu, Weiping Wang, Jie Zhou, Mo Yu</p><p><b>Upvotes:</b> 105</p><p><b>Summary:</b> Humans understand long and complex texts by relying on a holistic semantic representation of the content. This global view helps organize prior knowledge, interpret new information, and integrate evidence dispersed across a document, as revealed by the Mindscape-Aware Capability of humans in psychology. Current Retrieval-Augmented Generation (RAG) systems lack such guidance and therefore struggle with long-context tasks. In this paper, we propose Mindscape-Aware RAG (MiA-RAG), the first approach that equips LLM-based RAG systems with explicit global context awareness. MiA-RAG builds a mindscape through hierarchical summarization and conditions both retrieval and generation on this global semantic representation. This enables the retriever to form enriched query embeddings and the generator to reason over retrieved evidence within a coherent global context. We evaluate MiA-RAG across diverse long-context and bilingual benchmarks for evidence-based understanding and global sense-making. It consistently surpasses baselines, and further analysis shows that it aligns local details with a coherent global representation, enabling more human-like long-context retrieval and reasoning.</p> - https://arxiv.org/abs/2512.17220 - Fri, 19 Dec 2025 04:08:29 +0000 + IMA++: ISIC Archive Multi-Annotator Dermoscopic Skin Lesion Segmentation Dataset + https://arxiv.org/abs/2512.21472 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21472.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Kumar Abhishek, Jeremy Kawahara, Ghassan Hamarneh</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Multi-annotator medical image segmentation is an important research problem, but requires annotated datasets that are expensive to collect. Dermoscopic skin lesion imaging allows human experts and AI systems to observe morphological structures otherwise not discernable from regular clinical photographs. However, currently there are no large-scale publicly available multi-annotator skin lesion segmentation (SLS) datasets with annotator-labels for dermoscopic skin lesion imaging. We introduce ISIC MultiAnnot++, a large public multi-annotator skin lesion segmentation dataset for images from the ISIC Archive. The final dataset contains 17,684 segmentation masks spanning 14,967 dermoscopic images, where 2,394 dermoscopic images have 2-5 segmentations per image, making it the largest publicly available SLS dataset. Further, metadata about the segmentation, including the annotators' skill level and segmentation tool, is included, enabling research on topics such as annotator-specific preference modeling for segmentation and annotator metadata analysis. We provide an analysis on the characteristics of this dataset, curated data partitions, and consensus segmentation masks.</p> + https://arxiv.org/abs/2512.21472 + Thu, 25 Dec 2025 02:21:55 +0000 - InsertAnywhere: Bridging 4D Scene Geometry and Diffusion Models for Realistic Video Object Insertion - https://arxiv.org/abs/2512.17504 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.17504.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hoiyeong Jin, Hyojin Jang, Jeongho Kim, Junha Hyung, Kinam Kim, Dongjin Kim, Huijin Choi, Hyeonji Kim, Jaegul Choo</p><p><b>Upvotes:</b> 95</p><p><b>Summary:</b> Recent advances in diffusion-based video generation have opened new possibilities for controllable video editing, yet realistic video object insertion (VOI) remains challenging due to limited 4D scene understanding and inadequate handling of occlusion and lighting effects. We present InsertAnywhere, a new VOI framework that achieves geometrically consistent object placement and appearance-faithful video synthesis. Our method begins with a 4D aware mask generation module that reconstructs the scene geometry and propagates user specified object placement across frames while maintaining temporal coherence and occlusion consistency. Building upon this spatial foundation, we extend a diffusion based video generation model to jointly synthesize the inserted object and its surrounding local variations such as illumination and shading. To enable supervised training, we introduce ROSE++, an illumination aware synthetic dataset constructed by transforming the ROSE object removal dataset into triplets of object removed video, object present video, and a VLM generated reference image. Through extensive experiments, we demonstrate that our framework produces geometrically plausible and visually coherent object insertions across diverse real world scenarios, significantly outperforming existing research and commercial models.</p> - https://arxiv.org/abs/2512.17504 - Fri, 19 Dec 2025 12:14:36 +0000 + Can We Trust AI Explanations? Evidence of Systematic Underreporting in Chain-of-Thought Reasoning + https://arxiv.org/abs/2601.00830 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00830.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Deep Pankajbhai Mehta</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> When AI systems explain their reasoning step-by-step, practitioners often assume these explanations reveal what actually influenced the AI's answer. We tested this assumption by embedding hints into questions and measuring whether models mentioned them. In a study of over 9,000 test cases across 11 leading AI models, we found a troubling pattern: models almost never mention hints spontaneously, yet when asked directly, they admit noticing them. This suggests models see influential information but choose not to report it. Telling models they are being watched does not help. Forcing models to report hints works, but causes them to report hints even when none exist and reduces their accuracy. We also found that hints appealing to user preferences are especially dangerous-models follow them most often while reporting them least. These findings suggest that simply watching AI reasoning is not enough to catch hidden influences.</p> + https://arxiv.org/abs/2601.00830 + Thu, 25 Dec 2025 05:29:53 +0000 - InSight-o3: Empowering Multimodal Foundation Models with Generalized Visual Search - https://arxiv.org/abs/2512.18745 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.18745.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Kaican Li, Lewei Yao, Jiannan Wu, Tiezheng Yu, Jierun Chen, Haoli Bai, Lu Hou, Lanqing Hong, Wei Zhang, Nevin L. Zhang</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> The ability for AI agents to "think with images" requires a sophisticated blend of reasoning and perception. However, current open multimodal agents still largely fall short on the reasoning aspect crucial for real-world tasks like analyzing documents with dense charts/diagrams and navigating maps. To address this gap, we introduce O3-Bench, a new benchmark designed to evaluate multimodal reasoning with interleaved attention to visual details. O3-Bench features challenging problems that require agents to piece together subtle visual information from distinct image areas through multi-step reasoning. The problems are highly challenging even for frontier systems like OpenAI o3, which only obtains 40.8% accuracy on O3-Bench. To make progress, we propose InSight-o3, a multi-agent framework consisting of a visual reasoning agent (vReasoner) and a visual search agent (vSearcher) for which we introduce the task of generalized visual search -- locating relational, fuzzy, or conceptual regions described in free-form language, beyond just simple objects or figures in natural images. We then present a multimodal LLM purpose-trained for this task via reinforcement learning. As a plug-and-play agent, our vSearcher empowers frontier multimodal models (as vReasoners), significantly improving their performance on a wide range of benchmarks. This marks a concrete step towards powerful o3-like open systems. Our code and dataset can be found at https://github.com/m-Just/InSight-o3 .</p> - https://arxiv.org/abs/2512.18745 - Sun, 21 Dec 2025 14:23:07 +0000 + Few Tokens Matter: Entropy Guided Attacks on Vision-Language Models + https://arxiv.org/abs/2512.21815 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21815.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mengqi He, Xinyu Tian, Xin Shen, Jinhong Ni, Shu Zou, Zhaoyuan Yang, Jing Zhang</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Vision-language models (VLMs) achieve remarkable performance but remain vulnerable to adversarial attacks. Entropy, a measure of model uncertainty, is strongly correlated with the reliability of VLM. Prior entropy-based attacks maximize uncertainty at all decoding steps, implicitly assuming that every token contributes equally to generation instability. We show instead that a small fraction (about 20%) of high-entropy tokens, i.e., critical decision points in autoregressive generation, disproportionately governs output trajectories. By concentrating adversarial perturbations on these positions, we achieve semantic degradation comparable to global methods while using substantially smaller budgets. More importantly, across multiple representative VLMs, such selective attacks convert 35-49% of benign outputs into harmful ones, exposing a more critical safety risk. Remarkably, these vulnerable high-entropy forks recur across architecturally diverse VLMs, enabling feasible transferability (17-26% harmful rates on unseen targets). Motivated by these findings, we propose Entropy-bank Guided Adversarial attacks (EGA), which achieves competitive attack success rates (93-95%) alongside high harmful conversion, thereby revealing new weaknesses in current VLM safety mechanisms.</p> + https://arxiv.org/abs/2512.21815 + Fri, 26 Dec 2025 01:01:25 +0000 - CosineGate: Semantic Dynamic Routing via Cosine Incompatibility in Residual Networks - https://arxiv.org/abs/2512.22206 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22206.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yogeswar Reddy Thota</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Modern deep residual networks perform substantial redundant computation by evaluating all residual blocks for every input, even when identity mappings suffice. We introduce CosineGate, an end-to-end differentiable architecture for dynamic routing in residual networks that uses cosine incompatibility between identity and residual feature representations as a self-supervised skip signal. CosineGate measures semantic redundancy through the Cosine Incompatibility Ratio (CIR), defined as 1 - cos(x, F(x)), and uses Gumbel-Softmax relaxation to enable per-sample, per-block gating during training. A progressive FLOPs regularization term controls average compute usage without destabilizing optimization. On CIFAR-10, CosineGate spans the accuracy-efficiency Pareto frontier: an aggressive configuration achieves 89.9 percent accuracy with 24.1 percent FLOPs savings, a balanced configuration achieves 91.3 percent accuracy with 28.5 percent savings at epoch 160, and a conservative configuration reaches a peak of 93.2 percent accuracy with minimal compute reduction. These results match or exceed ResNet-20 (91.3 percent) while reducing computation, without auxiliary supervision, distillation, or task-specific heuristics. Our results demonstrate that simple geometric measures of feature incompatibility provide a principled and effective signal for dynamic residual routing.</p> - https://arxiv.org/abs/2512.22206 - Sun, 21 Dec 2025 18:26:18 +0000 + SciEvalKit: An Open-source Evaluation Toolkit for Scientific General Intelligence + https://arxiv.org/abs/2512.22334 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22334.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yiheng Wang, Yixin Chen, Shuo Li, Yifan Zhou, Bo Liu, Hengjian Gao, Jiakang Yuan, Jia Bu, Wanghan Xu, Yuhao Zhou, Xiangyu Zhao, Zhiwang Zhou, Fengxiang Wang, Haodong Duan, Songyang Zhang, Jun Yao, Han Deng, Yizhou Wang, Jiabei Xiao, Jiaqi Liu, Encheng Su, Yujie Liu, Weida Wang, Junchi Yao, Shenghe Zheng, Haoran Sun, Runmin Ma, Xiangchao Yan, Bo Zhang, Dongzhan Zhou, Shufei Zhang, Peng Ye, Xiaosong Wang, Shixiang Tang, Wenlong Zhang, Lei Bai</p><p><b>Upvotes:</b> 34</p><p><b>Summary:</b> We introduce SciEvalKit, a unified benchmarking toolkit designed to evaluate AI models for science across a broad range of scientific disciplines and task capabilities. Unlike general-purpose evaluation platforms, SciEvalKit focuses on the core competencies of scientific intelligence, including Scientific Multimodal Perception, Scientific Multimodal Reasoning, Scientific Multimodal Understanding, Scientific Symbolic Reasoning, Scientific Code Generation, Science Hypothesis Generation and Scientific Knowledge Understanding. It supports six major scientific domains, spanning from physics and chemistry to astronomy and materials science. SciEvalKit builds a foundation of expert-grade scientific benchmarks, curated from real-world, domain-specific datasets, ensuring that tasks reflect authentic scientific challenges. The toolkit features a flexible, extensible evaluation pipeline that enables batch evaluation across models and datasets, supports custom model and dataset integration, and provides transparent, reproducible, and comparable results. By bridging capability-based evaluation and disciplinary diversity, SciEvalKit offers a standardized yet customizable infrastructure to benchmark the next generation of scientific foundation models and intelligent agents. The toolkit is open-sourced and actively maintained to foster community-driven development and progress in AI4Science.</p> + https://arxiv.org/abs/2512.22334 + Fri, 26 Dec 2025 17:36:02 +0000 - DiRL: An Efficient Post-Training Framework for Diffusion Language Models - https://arxiv.org/abs/2512.22234 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22234.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ying Zhu, Jiaxin Wan, Xiaoran Liu, Siyanag He, Qiqi Wang, Xu Guo, Tianyi Liang, Zengfeng Huang, Ziwei He, Xipeng Qiu</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Diffusion Language Models (dLLMs) have emerged as promising alternatives to Auto-Regressive (AR) models. While recent efforts have validated their pre-training potential and accelerated inference speeds, the post-training landscape for dLLMs remains underdeveloped. Existing methods suffer from computational inefficiency and objective mismatches between training and inference, severely limiting performance on complex reasoning tasks such as mathematics. To address this, we introduce DiRL, an efficient post-training framework that tightly integrates FlexAttention-accelerated blockwise training with LMDeploy-optimized inference. This architecture enables a streamlined online model update loop, facilitating efficient two-stage post-training (Supervised Fine-Tuning followed by Reinforcement Learning). Building on this framework, we propose DiPO, the first unbiased Group Relative Policy Optimization (GRPO) implementation tailored for dLLMs. We validate our approach by training DiRL-8B-Instruct on high-quality math data. Our model achieves state-of-the-art math performance among dLLMs and surpasses comparable models in the Qwen2.5 series on several benchmarks.</p> - https://arxiv.org/abs/2512.22234 - Tue, 23 Dec 2025 08:33:19 +0000 + RGS-SLAM: Robust Gaussian Splatting SLAM with One-Shot Dense Initialization + https://arxiv.org/abs/2601.00705 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00705.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wei-Tse Cheng, Yen-Jen Chiou, Yuan-Fu Yang</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> We introduce RGS-SLAM, a robust Gaussian-splatting SLAM framework that replaces the residual-driven densification stage of GS-SLAM with a training-free correspondence-to-Gaussian initialization. Instead of progressively adding Gaussians as residuals reveal missing geometry, RGS-SLAM performs a one-shot triangulation of dense multi-view correspondences derived from DINOv3 descriptors refined through a confidence-aware inlier classifier, generating a well-distributed and structure-aware Gaussian seed prior to optimization. This initialization stabilizes early mapping and accelerates convergence by roughly 20\%, yielding higher rendering fidelity in texture-rich and cluttered scenes while remaining fully compatible with existing GS-SLAM pipelines. Evaluated on the TUM RGB-D and Replica datasets, RGS-SLAM achieves competitive or superior localization and reconstruction accuracy compared with state-of-the-art Gaussian and point-based SLAM systems, sustaining real-time mapping performance at up to 925 FPS.</p> + https://arxiv.org/abs/2601.00705 + Sun, 28 Dec 2025 03:45:57 +0000 - SlideTailor: Personalized Presentation Slide Generation for Scientific Papers - https://arxiv.org/abs/2512.20292 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20292.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wenzheng Zeng, Mingyu Ouyang, Langyuan Cui, Hwee Tou Ng</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Automatic presentation slide generation can greatly streamline content creation. However, since preferences of each user may vary, existing under-specified formulations often lead to suboptimal results that fail to align with individual user needs. We introduce a novel task that conditions paper-to-slides generation on user-specified preferences. We propose a human behavior-inspired agentic framework, SlideTailor, that progressively generates editable slides in a user-aligned manner. Instead of requiring users to write their preferences in detailed textual form, our system only asks for a paper-slides example pair and a visual template - natural and easy-to-provide artifacts that implicitly encode rich user preferences across content and visual style. Despite the implicit and unlabeled nature of these inputs, our framework effectively distills and generalizes the preferences to guide customized slide generation. We also introduce a novel chain-of-speech mechanism to align slide content with planned oral narration. Such a design significantly enhances the quality of generated slides and enables downstream applications like video presentations. To support this new task, we construct a benchmark dataset that captures diverse user preferences, with carefully designed interpretable metrics for robust evaluation. Extensive experiments demonstrate the effectiveness of our framework.</p> - https://arxiv.org/abs/2512.20292 - Tue, 23 Dec 2025 12:01:18 +0000 + Diversity or Precision? A Deep Dive into Next Token Prediction + https://arxiv.org/abs/2512.22955 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22955.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoyuan Wu, Hai Wang, Jiajia Wu, Jinxiang Ou, Keyao Wang, Weile Chen, Zihao Zheng, Bei Yu</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Recent advancements have shown that reinforcement learning (RL) can substantially improve the reasoning abilities of large language models (LLMs). The effectiveness of such RL training, however, depends critically on the exploration space defined by the pre-trained model's token-output distribution. In this paper, we revisit the standard cross-entropy loss, interpreting it as a specific instance of policy gradient optimization applied within a single-step episode. To systematically study how the pre-trained distribution shapes the exploration potential for subsequent RL, we propose a generalized pre-training objective that adapts on-policy RL principles to supervised learning. By framing next-token prediction as a stochastic decision process, we introduce a reward-shaping strategy that explicitly balances diversity and precision. Our method employs a positive reward scaling factor to control probability concentration on ground-truth tokens and a rank-aware mechanism that treats high-ranking and low-ranking negative tokens asymmetrically. This allows us to reshape the pre-trained token-output distribution and investigate how to provide a more favorable exploration space for RL, ultimately enhancing end-to-end reasoning performance. Contrary to the intuition that higher distribution entropy facilitates effective exploration, we find that imposing a precision-oriented prior yields a superior exploration space for RL.</p> + https://arxiv.org/abs/2512.22955 + Sun, 28 Dec 2025 14:53:24 +0000 - Masking Teacher and Reinforcing Student for Distilling Vision-Language Models - https://arxiv.org/abs/2512.22238 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22238.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Byung-Kwan Lee, Yu-Chiang Frank Wang, Ryo Hachiuma</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Large-scale vision-language models (VLMs) have recently achieved remarkable multimodal understanding, but their massive size makes them impractical for deployment on mobile or edge devices. This raises the need for compact yet capable VLMs that can efficiently learn from powerful large teachers. However, distilling knowledge from a large teacher to a small student remains challenging due to their large size gap: the student often fails to reproduce the teacher's complex, high-dimensional representations, leading to unstable learning and degraded performance. To address this, we propose Masters (Masking Teacher and Reinforcing Student), a mask-progressive reinforcement learning (RL) distillation framework. Masters first masks non-dominant weights of the teacher to reduce unnecessary complexity, then progressively restores the teacher by gradually increasing its capacity during training. This strategy allows the student to learn richer representations from the teacher in a smooth and stable manner. To further refine knowledge transfer, Masters integrates an offline RL stage with two complementary rewards: an accuracy reward that measures the correctness of the generated responses, and a distillation reward that quantifies the ease of transferring responses from teacher to student. Unlike online think-answer RL paradigms that are computationally expensive and generate lengthy responses, our offline RL leverages pre-generated responses from masked teachers. These provide rich yet efficient guidance, enabling students to achieve strong performance without requiring the think-answer process.</p> - https://arxiv.org/abs/2512.22238 - Tue, 23 Dec 2025 14:40:38 +0000 + Toward Stable Semi-Supervised Remote Sensing Segmentation via Co-Guidance and Co-Fusion + https://arxiv.org/abs/2512.23035 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23035.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yi Zhou, Xuechao Zou, Shun Zhang, Kai Li, Shiying Wang, Jingming Chen, Congyan Lang, Tengfei Cao, Pin Tao, Yuanchun Shi</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Semi-supervised remote sensing (RS) image semantic segmentation offers a promising solution to alleviate the burden of exhaustive annotation, yet it fundamentally struggles with pseudo-label drift, a phenomenon where confirmation bias leads to the accumulation of errors during training. In this work, we propose Co2S, a stable semi-supervised RS segmentation framework that synergistically fuses priors from vision-language models and self-supervised models. Specifically, we construct a heterogeneous dual-student architecture comprising two distinct ViT-based vision foundation models initialized with pretrained CLIP and DINOv3 to mitigate error accumulation and pseudo-label drift. To effectively incorporate these distinct priors, an explicit-implicit semantic co-guidance mechanism is introduced that utilizes text embeddings and learnable queries to provide explicit and implicit class-level guidance, respectively, thereby jointly enhancing semantic consistency. Furthermore, a global-local feature collaborative fusion strategy is developed to effectively fuse the global contextual information captured by CLIP with the local details produced by DINOv3, enabling the model to generate highly precise segmentation results. Extensive experiments on six popular datasets demonstrate the superiority of the proposed method, which consistently achieves leading performance across various partition protocols and diverse scenarios. Project page is available at https://xavierjiezou.github.io/Co2S/.</p> + https://arxiv.org/abs/2512.23035 + Sun, 28 Dec 2025 18:24:19 +0000 - Quantile Rendering: Efficiently Embedding High-dimensional Feature on 3D Gaussian Splatting - https://arxiv.org/abs/2512.20927 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20927.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yoonwoo Jeong, Cheng Sun, Frank Wang, Minsu Cho, Jaesung Choe</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Recent advancements in computer vision have successfully extended Open-vocabulary segmentation (OVS) to the 3D domain by leveraging 3D Gaussian Splatting (3D-GS). Despite this progress, efficiently rendering the high-dimensional features required for open-vocabulary queries poses a significant challenge. Existing methods employ codebooks or feature compression, causing information loss, thereby degrading segmentation quality. To address this limitation, we introduce Quantile Rendering (Q-Render), a novel rendering strategy for 3D Gaussians that efficiently handles high-dimensional features while maintaining high fidelity. Unlike conventional volume rendering, which densely samples all 3D Gaussians intersecting each ray, Q-Render sparsely samples only those with dominant influence along the ray. By integrating Q-Render into a generalizable 3D neural network, we also propose Gaussian Splatting Network (GS-Net), which predicts Gaussian features in a generalizable manner. Extensive experiments on ScanNet and LeRF demonstrate that our framework outperforms state-of-the-art methods, while enabling real-time rendering with an approximate ~43.7x speedup on 512-D feature maps. Code will be made publicly available.</p> - https://arxiv.org/abs/2512.20927 - Wed, 24 Dec 2025 04:16:18 +0000 + MindWatcher: Toward Smarter Multimodal Tool-Integrated Reasoning + https://arxiv.org/abs/2512.23412 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23412.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiawei Chen, Xintian Shen, Lihao Zheng, Zhenwei Shao, Hongyuan Zhang, Pengfei Yu, Xudong Rao, Ning Mao, Xiaobo Liu, Lian Wen, Chaoqun Du, Feng Gu, Wei He, Qizhen Li, Shanshan Li, Zide Liu, Jing Luo, Lifu Mu, Xuhao Pan, Chang Ren, Haoyi Sun, Qian Wang, Wei Wang, Hongfu Yang, Jiqing Zhan, Chunpeng Zhou, Zheng Zhou, Hao Ma, Tao Wei, Pan Zhou, Wei Chen</p><p><b>Upvotes:</b> 36</p><p><b>Summary:</b> Traditional workflow-based agents exhibit limited intelligence when addressing real-world problems requiring tool invocation. Tool-integrated reasoning (TIR) agents capable of autonomous reasoning and tool invocation are rapidly emerging as a powerful approach for complex decision-making tasks involving multi-step interactions with external environments. In this work, we introduce MindWatcher, a TIR agent integrating interleaved thinking and multimodal chain-of-thought (CoT) reasoning. MindWatcher can autonomously decide whether and how to invoke diverse tools and coordinate their use, without relying on human prompts or workflows. The interleaved thinking paradigm enables the model to switch between thinking and tool calling at any intermediate stage, while its multimodal CoT capability allows manipulation of images during reasoning to yield more precise search results. We implement automated data auditing and evaluation pipelines, complemented by manually curated high-quality datasets for training, and we construct a benchmark, called MindWatcher-Evaluate Bench (MWE-Bench), to evaluate its performance. MindWatcher is equipped with a comprehensive suite of auxiliary reasoning tools, enabling it to address broad-domain multimodal problems. A large-scale, high-quality local image retrieval database, covering eight categories including cars, animals, and plants, endows model with robust object recognition despite its small size. Finally, we design a more efficient training infrastructure for MindWatcher, enhancing training speed and hardware utilization. Experiments not only demonstrate that MindWatcher matches or exceeds the performance of larger or more recent models through superior tool invocation, but also uncover critical insights for agent training, such as the genetic inheritance phenomenon in agentic RL.</p> + https://arxiv.org/abs/2512.23412 + Mon, 29 Dec 2025 12:16:12 +0000 - GateBreaker: Gate-Guided Attacks on Mixture-of-Expert LLMs - https://arxiv.org/abs/2512.21008 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21008.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Lichao Wu, Sasha Behrouzi, Mohamadreza Rostami, Stjepan Picek, Ahmad-Reza Sadeghi</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Mixture-of-Experts (MoE) architectures have advanced the scaling of Large Language Models (LLMs) by activating only a sparse subset of parameters per input, enabling state-of-the-art performance with reduced computational cost. As these models are increasingly deployed in critical domains, understanding and strengthening their alignment mechanisms is essential to prevent harmful outputs. However, existing LLM safety research has focused almost exclusively on dense architectures, leaving the unique safety properties of MoEs largely unexamined. The modular, sparsely-activated design of MoEs suggests that safety mechanisms may operate differently than in dense models, raising questions about their robustness. In this paper, we present GateBreaker, the first training-free, lightweight, and architecture-agnostic attack framework that compromises the safety alignment of modern MoE LLMs at inference time. GateBreaker operates in three stages: (i) gate-level profiling, which identifies safety experts disproportionately routed on harmful inputs, (ii) expert-level localization, which localizes the safety structure within safety experts, and (iii) targeted safety removal, which disables the identified safety structure to compromise the safety alignment. Our study shows that MoE safety concentrates within a small subset of neurons coordinated by sparse routing. Selective disabling of these neurons, approximately 3% of neurons in the targeted expert layers, significantly increases the averaged attack success rate (ASR) from 7.4% to 64.9% against the eight latest aligned MoE LLMs with limited utility degradation. These safety neurons transfer across models within the same family, raising ASR from 17.9% to 67.7% with one-shot transfer attack. Furthermore, GateBreaker generalizes to five MoE vision language models (VLMs) with 60.9% ASR on unsafe image inputs.</p> - https://arxiv.org/abs/2512.21008 - Wed, 24 Dec 2025 07:13:24 +0000 + Memorization in 3D Shape Generation: An Empirical Study + https://arxiv.org/abs/2512.23628 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23628.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shu Pu, Boya Zeng, Kaichen Zhou, Mengyu Wang, Zhuang Liu</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Generative models are increasingly used in 3D vision to synthesize novel shapes, yet it remains unclear whether their generation relies on memorizing training shapes. Understanding their memorization could help prevent training data leakage and improve the diversity of generated results. In this paper, we design an evaluation framework to quantify memorization in 3D generative models and study the influence of different data and modeling designs on memorization. We first apply our framework to quantify memorization in existing methods. Next, through controlled experiments with a latent vector-set (Vecset) diffusion model, we find that, on the data side, memorization depends on data modality, and increases with data diversity and finer-grained conditioning; on the modeling side, it peaks at a moderate guidance scale and can be mitigated by longer Vecsets and simple rotation augmentation. Together, our framework and analysis provide an empirical understanding of memorization in 3D generative models and suggest simple yet effective strategies to reduce it without degrading generation quality. Our code is available at https://github.com/zlab-princeton/3d_mem.</p> + https://arxiv.org/abs/2512.23628 + Mon, 29 Dec 2025 17:39:21 +0000 - Shape of Thought: When Distribution Matters More than Correctness in Reasoning Tasks - https://arxiv.org/abs/2512.22255 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22255.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Abhranil Chandra, Ayush Agrawal, Arian Hosseini, Sebastian Fischmeister, Rishabh Agarwal, Navin Goyal, Aaron Courville</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> We present the surprising finding that a language model's reasoning capabilities can be improved by training on synthetic datasets of chain-of-thought (CoT) traces from more capable models, even when all of those traces lead to an incorrect final answer. Our experiments show this approach can yield better performance on reasoning tasks than training on human-annotated datasets. We hypothesize that two key factors explain this phenomenon: first, the distribution of synthetic data is inherently closer to the language model's own distribution, making it more amenable to learning. Second, these `incorrect' traces are often only partially flawed and contain valid reasoning steps from which the model can learn. To further test the first hypothesis, we use a language model to paraphrase human-annotated traces -- shifting their distribution closer to the model's own distribution -- and show that this improves performance. For the second hypothesis, we introduce increasingly flawed CoT traces and study to what extent models are tolerant to these flaws. We demonstrate our findings across various reasoning domains like math, algorithmic reasoning and code generation using MATH, GSM8K, Countdown and MBPP datasets on various language models ranging from 1.5B to 9B across Qwen, Llama, and Gemma models. Our study shows that curating datasets that are closer to the model's distribution is a critical aspect to consider. We also show that a correct final answer is not always a reliable indicator of a faithful reasoning process.</p> - https://arxiv.org/abs/2512.22255 - Wed, 24 Dec 2025 07:35:55 +0000 + GARDO: Reinforcing Diffusion Models without Reward Hacking + https://arxiv.org/abs/2512.24138 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24138.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoran He, Yuxiao Ye, Jie Liu, Jiajun Liang, Zhiyong Wang, Ziyang Yuan, Xintao Wang, Hangyu Mao, Pengfei Wan, Ling Pan</p><p><b>Upvotes:</b> 28</p><p><b>Summary:</b> Fine-tuning diffusion models via online reinforcement learning (RL) has shown great potential for enhancing text-to-image alignment. However, since precisely specifying a ground-truth objective for visual tasks remains challenging, the models are often optimized using a proxy reward that only partially captures the true goal. This mismatch often leads to reward hacking, where proxy scores increase while real image quality deteriorates and generation diversity collapses. While common solutions add regularization against the reference policy to prevent reward hacking, they compromise sample efficiency and impede the exploration of novel, high-reward regions, as the reference policy is usually sub-optimal. To address the competing demands of sample efficiency, effective exploration, and mitigation of reward hacking, we propose Gated and Adaptive Regularization with Diversity-aware Optimization (GARDO), a versatile framework compatible with various RL algorithms. Our key insight is that regularization need not be applied universally; instead, it is highly effective to selectively penalize a subset of samples that exhibit high uncertainty. To address the exploration challenge, GARDO introduces an adaptive regularization mechanism wherein the reference model is periodically updated to match the capabilities of the online policy, ensuring a relevant regularization target. To address the mode collapse issue in RL, GARDO amplifies the rewards for high-quality samples that also exhibit high diversity, encouraging mode coverage without destabilizing the optimization process. Extensive experiments across diverse proxy rewards and hold-out unseen metrics consistently show that GARDO mitigates reward hacking and enhances generation diversity without sacrificing sample efficiency or exploration, highlighting its effectiveness and robustness.</p> + https://arxiv.org/abs/2512.24138 + Tue, 30 Dec 2025 10:55:45 +0000 - UltraShape 1.0: High-Fidelity 3D Shape Generation via Scalable Geometric Refinement - https://arxiv.org/abs/2512.21185 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21185.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tanghui Jia, Dongyu Yan, Dehao Hao, Yang Li, Kaiyi Zhang, Xianyi He, Lanjiong Li, Jinnan Chen, Lutao Jiang, Qishen Yin, Long Quan, Ying-Cong Chen, Li Yuan</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> In this report, we introduce UltraShape 1.0, a scalable 3D diffusion framework for high-fidelity 3D geometry generation. The proposed approach adopts a two-stage generation pipeline: a coarse global structure is first synthesized and then refined to produce detailed, high-quality geometry. To support reliable 3D generation, we develop a comprehensive data processing pipeline that includes a novel watertight processing method and high-quality data filtering. This pipeline improves the geometric quality of publicly available 3D datasets by removing low-quality samples, filling holes, and thickening thin structures, while preserving fine-grained geometric details. To enable fine-grained geometry refinement, we decouple spatial localization from geometric detail synthesis in the diffusion process. We achieve this by performing voxel-based refinement at fixed spatial locations, where voxel queries derived from coarse geometry provide explicit positional anchors encoded via RoPE, allowing the diffusion model to focus on synthesizing local geometric details within a reduced, structured solution space. Our model is trained exclusively on publicly available 3D datasets, achieving strong geometric quality despite limited training resources. Extensive evaluations demonstrate that UltraShape 1.0 performs competitively with existing open-source methods in both data processing quality and geometry generation. All code and trained models will be released to support future research.</p> - https://arxiv.org/abs/2512.21185 - Wed, 24 Dec 2025 14:08:38 +0000 + Taming Preference Mode Collapse via Directional Decoupling Alignment in Diffusion Reinforcement Learning + https://arxiv.org/abs/2512.24146 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24146.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chubin Chen, Sujie Hu, Jiashu Zhu, Meiqi Wu, Jintao Chen, Yanxun Li, Nisha Huang, Chengyu Fang, Jiahong Wu, Xiangxiang Chu, Xiu Li</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Recent studies have demonstrated significant progress in aligning text-to-image diffusion models with human preference via Reinforcement Learning from Human Feedback. However, while existing methods achieve high scores on automated reward metrics, they often lead to Preference Mode Collapse (PMC)-a specific form of reward hacking where models converge on narrow, high-scoring outputs (e.g., images with monolithic styles or pervasive overexposure), severely degrading generative diversity. In this work, we introduce and quantify this phenomenon, proposing DivGenBench, a novel benchmark designed to measure the extent of PMC. We posit that this collapse is driven by over-optimization along the reward model's inherent biases. Building on this analysis, we propose Directional Decoupling Alignment (D^2-Align), a novel framework that mitigates PMC by directionally correcting the reward signal. Specifically, our method first learns a directional correction within the reward model's embedding space while keeping the model frozen. This correction is then applied to the reward signal during the optimization process, preventing the model from collapsing into specific modes and thereby maintaining diversity. Our comprehensive evaluation, combining qualitative analysis with quantitative metrics for both quality and diversity, reveals that D^2-Align achieves superior alignment with human preference.</p> + https://arxiv.org/abs/2512.24146 + Tue, 30 Dec 2025 11:17:52 +0000 - SVBench: Evaluation of Video Generation Models on Social Reasoning - https://arxiv.org/abs/2512.21507 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21507.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wenshuo Peng, Gongxuan Wang, Tianmeng Yang, Chuanhao Li, Xiaojie Xu, Hui He, Kaipeng Zhang</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Recent text-to-video generation models exhibit remarkable progress in visual realism, motion fidelity, and text-video alignment, yet they remain fundamentally limited in their ability to generate socially coherent behavior. Unlike humans, who effortlessly infer intentions, beliefs, emotions, and social norms from brief visual cues, current models tend to render literal scenes without capturing the underlying causal or psychological logic. To systematically evaluate this gap, we introduce the first benchmark for social reasoning in video generation. Grounded in findings from developmental and social psychology, our benchmark organizes thirty classic social cognition paradigms into seven core dimensions, including mental-state inference, goal-directed action, joint attention, social coordination, prosocial behavior, social norms, and multi-agent strategy. To operationalize these paradigms, we develop a fully training-free agent-based pipeline that (i) distills the reasoning mechanism of each experiment, (ii) synthesizes diverse video-ready scenarios, (iii) enforces conceptual neutrality and difficulty control through cue-based critique, and (iv) evaluates generated videos using a high-capacity VLM judge across five interpretable dimensions of social reasoning. Using this framework, we conduct the first large-scale study across seven state-of-the-art video generation systems. Our results reveal substantial performance gaps: while modern models excel in surface-level plausibility, they systematically fail in intention recognition, belief reasoning, joint attention, and prosocial inference.</p> - https://arxiv.org/abs/2512.21507 - Thu, 25 Dec 2025 04:44:59 +0000 + Towards Open-Vocabulary Industrial Defect Understanding with a Large-Scale Multimodal Dataset + https://arxiv.org/abs/2512.24160 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24160.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> TsaiChing Ni, ZhenQi Chen, YuanFu Yang</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> We present IMDD-1M, the first large-scale Industrial Multimodal Defect Dataset comprising 1,000,000 aligned image-text pairs, designed to advance multimodal learning for manufacturing and quality inspection. IMDD-1M contains high-resolution real-world defects spanning over 60 material categories and more than 400 defect types, each accompanied by expert-verified annotations and fine-grained textual descriptions detailing defect location, severity, and contextual attributes. This dataset enables a wide spectrum of applications, including classification, segmentation, retrieval, captioning, and generative modeling. Building upon IMDD-1M, we train a diffusion-based vision-language foundation model from scratch, specifically tailored for industrial scenarios. The model serves as a generalizable foundation that can be efficiently adapted to specialized domains through lightweight fine-tuning. With less than 5% of the task-specific data required by dedicated expert models, it achieves comparable performance, highlighting the potential of data-efficient foundation model adaptation for industrial inspection and generation, paving the way for scalable, domain-adaptive, and knowledge-grounded manufacturing intelligence.</p> + https://arxiv.org/abs/2512.24160 + Tue, 30 Dec 2025 11:45:22 +0000 - Valori: A Deterministic Memory Substrate for AI Systems - https://arxiv.org/abs/2512.22280 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22280.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Varshith Gudur</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Modern AI systems rely on vector embeddings stored and searched using floating-point arithmetic. While effective for approximate similarity search, this design introduces fundamental non-determinism: identical models, inputs, and code can produce different memory states and retrieval results across hardware architectures (e.g., x86 vs. ARM). This prevents replayability and safe deployment, leading to silent data divergence that prevents post-hoc verification and compromises audit trails in regulated sectors. We present Valori, a deterministic AI memory substrate that replaces floating-point memory operations with fixed-point arithmetic (Q16.16) and models memory as a replayable state machine. Valori guarantees bit-identical memory states, snapshots, and search results across platforms. We demonstrate that non-determinism arises before indexing or retrieval and show how Valori enforces determinism at the memory boundary. Our results suggest that deterministic memory is a necessary primitive for trustworthy AI systems. The reference implementation is open-source and available at https://github.com/varshith-Git/Valori-Kernel (archived at https://zenodo.org/records/18022660).</p> - https://arxiv.org/abs/2512.22280 - Thu, 25 Dec 2025 06:04:04 +0000 + Taming Hallucinations: Boosting MLLMs' Video Understanding via Counterfactual Video Generation + https://arxiv.org/abs/2512.24271 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24271.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhe Huang, Hao Wen, Aiming Hao, Bingze Song, Meiqi Wu, Jiahong Wu, Xiangxiang Chu, Sheng Lu, Haoqian Wang</p><p><b>Upvotes:</b> 50</p><p><b>Summary:</b> Multimodal Large Language Models (MLLMs) have made remarkable progress in video understanding. However, they suffer from a critical vulnerability: an over-reliance on language priors, which can lead to visual ungrounded hallucinations, especially when processing counterfactual videos that defy common sense. This limitation, stemming from the intrinsic data imbalance between text and video, is challenging to address due to the substantial cost of collecting and annotating counterfactual data. To address this, we introduce DualityForge, a novel counterfactual data synthesis framework that employs controllable, diffusion-based video editing to transform real-world videos into counterfactual scenarios. By embedding structured contextual information into the video editing and QA generation processes, the framework automatically produces high-quality QA pairs together with original-edited video pairs for contrastive training. Based on this, we build DualityVidQA, a large-scale video dataset designed to reduce MLLM hallucinations. In addition, to fully exploit the contrastive nature of our paired data, we propose Duality-Normalized Advantage Training (DNA-Train), a two-stage SFT-RL training regime where the RL phase applies pair-wise ell_1 advantage normalization, thereby enabling a more stable and efficient policy optimization. Experiments on DualityVidQA-Test demonstrate that our method substantially reduces model hallucinations on counterfactual videos, yielding a relative improvement of 24.0% over the Qwen2.5-VL-7B baseline. Moreover, our approach achieves significant gains across both hallucination and general-purpose benchmarks, indicating strong generalization capability. We will open-source our dataset and code.</p> + https://arxiv.org/abs/2512.24271 + Tue, 30 Dec 2025 14:53:33 +0000 - Rethinking Sample Polarity in Reinforcement Learning with Verifiable Rewards - https://arxiv.org/abs/2512.21625 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21625.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xinyu Tang, Yuliang Zhan, Zhixun Li, Wayne Xin Zhao, Zhenduo Zhang, Zujie Wen, Zhiqiang Zhang, Jun Zhou</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Large reasoning models (LRMs) are typically trained using reinforcement learning with verifiable reward (RLVR) to enhance their reasoning abilities. In this paradigm, policies are updated using both positive and negative self-generated rollouts, which correspond to distinct sample polarities. In this paper, we provide a systematic investigation into how these sample polarities affect RLVR training dynamics and behaviors. We find that positive samples sharpen existing correct reasoning patterns, while negative samples encourage exploration of new reasoning paths. We further explore how adjusting the advantage values of positive and negative samples at both the sample level and the token level affects RLVR training. Based on these insights, we propose an Adaptive and Asymmetric token-level Advantage shaping method for Policy Optimization, namely A3PO, that more precisely allocates advantage signals to key tokens across different polarities. Experiments across five reasoning benchmarks demonstrate the effectiveness of our approach.</p> - https://arxiv.org/abs/2512.21625 - Thu, 25 Dec 2025 11:15:46 +0000 + SenseNova-MARS: Empowering Multimodal Agentic Reasoning and Search via Reinforcement Learning + https://arxiv.org/abs/2512.24330 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24330.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yong Xien Chng, Tao Hu, Wenwen Tong, Xueheng Li, Jiandong Chen, Haojia Yu, Jiefan Lu, Hewei Guo, Hanming Deng, Chengjun Xie, Gao Huang, Dahua Lin, Lewei Lu</p><p><b>Upvotes:</b> 33</p><p><b>Summary:</b> While Vision-Language Models (VLMs) can solve complex tasks through agentic reasoning, their capabilities remain largely constrained to text-oriented chain-of-thought or isolated tool invocation. They fail to exhibit the human-like proficiency required to seamlessly interleave dynamic tool manipulation with continuous reasoning, particularly in knowledge-intensive and visually complex scenarios that demand coordinated external tools such as search and image cropping. In this work, we introduce SenseNova-MARS, a novel Multimodal Agentic Reasoning and Search framework that empowers VLMs with interleaved visual reasoning and tool-use capabilities via reinforcement learning (RL). Specifically, SenseNova-MARS dynamically integrates the image search, text search, and image crop tools to tackle fine-grained and knowledge-intensive visual understanding challenges. In the RL stage, we propose the Batch-Normalized Group Sequence Policy Optimization (BN-GSPO) algorithm to improve the training stability and advance the model's ability to invoke tools and reason effectively. To comprehensively evaluate the agentic VLMs on complex visual tasks, we introduce the HR-MMSearch benchmark, the first search-oriented benchmark composed of high-resolution images with knowledge-intensive and search-driven questions. Experiments demonstrate that SenseNova-MARS achieves state-of-the-art performance on open-source search and fine-grained image understanding benchmarks. Specifically, on search-oriented benchmarks, SenseNova-MARS-8B scores 67.84 on MMSearch and 41.64 on HR-MMSearch, surpassing proprietary models such as Gemini-3-Flash and GPT-5. SenseNova-MARS represents a promising step toward agentic VLMs by providing effective and robust tool-use capabilities. To facilitate further research in this field, we will release all code, models, and datasets.</p> + https://arxiv.org/abs/2512.24330 + Tue, 30 Dec 2025 16:31:45 +0000 - Omni-Weather: Unified Multimodal Foundation Model for Weather Generation and Understanding - https://arxiv.org/abs/2512.21643 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21643.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhiwang Zhou, Yuandong Pu, Xuming He, Yidi Liu, Yixin Chen, Junchao Gong, Xiang Zhuang, Wanghan Xu, Qinglong Cao, Shixiang Tang, Yihao Liu, Wenlong Zhang, Lei Bai</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Weather modeling requires both accurate prediction and mechanistic interpretation, yet existing methods treat these goals in isolation, separating generation from understanding. To address this gap, we present Omni-Weather, the first multimodal foundation model that unifies weather generation and understanding within a single architecture. Omni-Weather integrates a radar encoder for weather generation tasks, followed by unified processing using a shared self-attention mechanism. Moreover, we construct a Chain-of-Thought dataset for causal reasoning in weather generation, enabling interpretable outputs and improved perceptual quality. Extensive experiments show Omni-Weather achieves state-of-the-art performance in both weather generation and understanding. Our findings further indicate that generative and understanding tasks in the weather domain can mutually enhance each other. Omni-Weather also demonstrates the feasibility and value of unifying weather generation and understanding.</p> - https://arxiv.org/abs/2512.21643 - Thu, 25 Dec 2025 12:08:09 +0000 + Recursive Language Models + https://arxiv.org/abs/2512.24601 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24601.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Alex L. Zhang, Tim Kraska, Omar Khattab</p><p><b>Upvotes:</b> 50</p><p><b>Summary:</b> We study allowing large language models (LLMs) to process arbitrarily long prompts through the lens of inference-time scaling. We propose Recursive Language Models (RLMs), a general inference strategy that treats long prompts as part of an external environment and allows the LLM to programmatically examine, decompose, and recursively call itself over snippets of the prompt. We find that RLMs successfully handle inputs up to two orders of magnitude beyond model context windows and, even for shorter prompts, dramatically outperform the quality of base LLMs and common long-context scaffolds across four diverse long-context tasks, while having comparable (or cheaper) cost per query.</p> + https://arxiv.org/abs/2512.24601 + Wed, 31 Dec 2025 03:43:41 +0000 - UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture - https://arxiv.org/abs/2512.21675 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21675.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuo Cao, Jiayang Li, Xiaohui Li, Yuandong Pu, Kaiwen Zhu, Yuanting Gao, Siqi Luo, Yi Xin, Qi Qin, Yu Zhou, Xiangyu Chen, Wenlong Zhang, Bin Fu, Yu Qiao, Yihao Liu</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> Multimodal large language models (MLLMs) have achieved remarkable progress in visual understanding tasks such as visual grounding, segmentation, and captioning. However, their ability to perceive perceptual-level image features remains limited. In this work, we present UniPercept-Bench, a unified framework for perceptual-level image understanding across three key domains: Aesthetics, Quality, Structure and Texture. We establish a hierarchical definition system and construct large-scale datasets to evaluate perceptual-level image understanding. Based on this foundation, we develop a strong baseline UniPercept trained via Domain-Adaptive Pre-Training and Task-Aligned RL, enabling robust generalization across both Visual Rating (VR) and Visual Question Answering (VQA) tasks. UniPercept outperforms existing MLLMs on perceptual-level image understanding and can serve as a plug-and-play reward model for text-to-image generation. This work defines Perceptual-Level Image Understanding in the era of MLLMs and, through the introduction of a comprehensive benchmark together with a strong baseline, provides a solid foundation for advancing perceptual-level multimodal image understanding.</p> - https://arxiv.org/abs/2512.21675 - Thu, 25 Dec 2025 13:35:52 +0000 + Youtu-Agent: Scaling Agent Productivity with Automated Generation and Hybrid Policy Optimization + https://arxiv.org/abs/2512.24615 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24615.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuchen Shi, Yuzheng Cai, Siqi Cai, Zihan Xu, Lichao Chen, Yulei Qin, Zhijian Zhou, Xiang Fei, Chaofan Qiu, Xiaoyu Tan, Gang Li, Zongyi Li, Haojia Lin, Guocan Cai, Yong Mao, Yunsheng Wu, Ke Li, Xing Sun</p><p><b>Upvotes:</b> 108</p><p><b>Summary:</b> Existing Large Language Model (LLM) agent frameworks face two significant challenges: high configuration costs and static capabilities. Building a high-quality agent often requires extensive manual effort in tool integration and prompt engineering, while deployed agents struggle to adapt to dynamic environments without expensive fine-tuning. To address these issues, we propose Youtu-Agent, a modular framework designed for the automated generation and continuous evolution of LLM agents. Youtu-Agent features a structured configuration system that decouples execution environments, toolkits, and context management, enabling flexible reuse and automated synthesis. We introduce two generation paradigms: a Workflow mode for standard tasks and a Meta-Agent mode for complex, non-standard requirements, capable of automatically generating tool code, prompts, and configurations. Furthermore, Youtu-Agent establishes a hybrid policy optimization system: (1) an Agent Practice module that enables agents to accumulate experience and improve performance through in-context optimization without parameter updates; and (2) an Agent RL module that integrates with distributed training frameworks to enable scalable and stable reinforcement learning of any Youtu-Agents in an end-to-end, large-scale manner. Experiments demonstrate that Youtu-Agent achieves state-of-the-art performance on WebWalkerQA (71.47\%) and GAIA (72.8\%) using open-weight models. Our automated generation pipeline achieves over 81\% tool synthesis success rate, while the Practice module improves performance on AIME 2024/2025 by +2.7\% and +5.4\% respectively. Moreover, our Agent RL training achieves 40\% speedup with steady performance improvement on 7B LLMs, enhancing coding/reasoning and searching capabilities respectively up to 35\% and 21\% on Maths and general/multi-hop QA benchmarks.</p> + https://arxiv.org/abs/2512.24615 + Wed, 31 Dec 2025 04:17:36 +0000 - An Information Theoretic Perspective on Agentic System Design - https://arxiv.org/abs/2512.21720 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21720.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shizhe He, Avanika Narayan, Ishan S. Khare, Scott W. Linderman, Christopher Ré, Dan Biderman</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Agentic language model (LM) systems power modern applications like "Deep Research" and "Claude Code," and leverage multi-LM architectures to overcome context limitations. Beneath their apparent diversity lies a recurring pattern: smaller "compressor" LMs (that can even run locally) distill raw context into compact text that is then consumed by larger "predictor" LMs. Despite their popularity, the design of compressor-predictor systems remains largely ad hoc, with little guidance on how compressor and predictor choices shape downstream performance. In practice, attributing gains to compression versus prediction requires costly, task-specific pairwise sweeps. We argue that these agentic system design questions are, at root, information-theoretic. Viewing the compressor LM as a noisy channel, we introduce a simple estimator of mutual information between the context and its compression to quantify compression quality in a task-independent way. We show that mutual information strongly predicts downstream performance, independent of any specific task. Through an information-theoretic framework, we perform a comprehensive empirical analysis across five datasets and three model families. Results reveal that larger compressors not only are more accurate, but also more token-efficient, conveying more bits of information per token. A 7B Qwen-2.5 compressor, for instance, is 1.6times more accurate, 4.6times more concise, and conveys 5.5times more bits of mutual information per token than its 1.5B sibling. Across datasets, scaling compressors is substantially more effective than scaling predictors, enabling larger on-device compressors to pair with smaller cloud predictors. Applied to a Deep Research system, these principles enable local compressors as small as 3B parameters to recover 99% of frontier-LM accuracy at 26% of API costs.</p> - https://arxiv.org/abs/2512.21720 - Thu, 25 Dec 2025 15:45:31 +0000 + Nested Learning: The Illusion of Deep Learning Architectures + https://arxiv.org/abs/2512.24695 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24695.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ali Behrouz, Meisam Razaviyayn, Peilin Zhong, Vahab Mirrokni</p><p><b>Upvotes:</b> 34</p><p><b>Summary:</b> Despite the recent progresses, particularly in developing Language Models, there are fundamental challenges and unanswered questions about how such models can continually learn/memorize, self-improve, and find effective solutions. In this paper, we present a new learning paradigm, called Nested Learning (NL), that coherently represents a machine learning model with a set of nested, multi-level, and/or parallel optimization problems, each of which with its own context flow. Through the lenses of NL, existing deep learning methods learns from data through compressing their own context flow, and in-context learning naturally emerges in large models. NL suggests a philosophy to design more expressive learning algorithms with more levels, resulting in higher-order in-context learning and potentially unlocking effective continual learning capabilities. We advocate for NL by presenting three core contributions: (1) Expressive Optimizers: We show that known gradient-based optimizers, such as Adam, SGD with Momentum, etc., are in fact associative memory modules that aim to compress the gradients' information (by gradient descent). Building on this insight, we present other more expressive optimizers with deep memory and/or more powerful learning rules; (2) Self-Modifying Learning Module: Taking advantage of NL's insights on learning algorithms, we present a sequence model that learns how to modify itself by learning its own update algorithm; and (3) Continuum Memory System: We present a new formulation for memory system that generalizes the traditional viewpoint of long/short-term memory. Combining our self-modifying sequence model with the continuum memory system, we present a continual learning module, called Hope, showing promising results in language modeling, knowledge incorporation, and few-shot generalization tasks, continual learning, and long-context reasoning tasks.</p> + https://arxiv.org/abs/2512.24695 + Wed, 31 Dec 2025 07:59:43 +0000 - Knot Forcing: Taming Autoregressive Video Diffusion Models for Real-time Infinite Interactive Portrait Animation - https://arxiv.org/abs/2512.21734 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21734.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Steven Xiao, Xindi Zhang, Dechao Meng, Qi Wang, Peng Zhang, Bang Zhang</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Real-time portrait animation is essential for interactive applications such as virtual assistants and live avatars, requiring high visual fidelity, temporal coherence, ultra-low latency, and responsive control from dynamic inputs like reference images and driving signals. While diffusion-based models achieve strong quality, their non-causal nature hinders streaming deployment. Causal autoregressive video generation approaches enable efficient frame-by-frame generation but suffer from error accumulation, motion discontinuities at chunk boundaries, and degraded long-term consistency. In this work, we present a novel streaming framework named Knot Forcing for real-time portrait animation that addresses these challenges through three key designs: (1) a chunk-wise generation strategy with global identity preservation via cached KV states of the reference image and local temporal modeling using sliding window attention; (2) a temporal knot module that overlaps adjacent chunks and propagates spatio-temporal cues via image-to-video conditioning to smooth inter-chunk motion transitions; and (3) A "running ahead" mechanism that dynamically updates the reference frame's temporal coordinate during inference, keeping its semantic context ahead of the current rollout frame to support long-term coherence. Knot Forcing enables high-fidelity, temporally consistent, and interactive portrait animation over infinite sequences, achieving real-time performance with strong visual stability on consumer-grade GPUs.</p> - https://arxiv.org/abs/2512.21734 - Thu, 25 Dec 2025 16:34:56 +0000 + MorphAny3D: Unleashing the Power of Structured Latent in 3D Morphing + https://arxiv.org/abs/2601.00204 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00204.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiaokun Sun, Zeyu Cai, Hao Tang, Ying Tai, Jian Yang, Zhenyu Zhang</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> 3D morphing remains challenging due to the difficulty of generating semantically consistent and temporally smooth deformations, especially across categories. We present MorphAny3D, a training-free framework that leverages Structured Latent (SLAT) representations for high-quality 3D morphing. Our key insight is that intelligently blending source and target SLAT features within the attention mechanisms of 3D generators naturally produces plausible morphing sequences. To this end, we introduce Morphing Cross-Attention (MCA), which fuses source and target information for structural coherence, and Temporal-Fused Self-Attention (TFSA), which enhances temporal consistency by incorporating features from preceding frames. An orientation correction strategy further mitigates the pose ambiguity within the morphing steps. Extensive experiments show that our method generates state-of-the-art morphing sequences, even for challenging cross-category cases. MorphAny3D further supports advanced applications such as decoupled morphing and 3D style transfer, and can be generalized to other SLAT-based generative models. Project page: https://xiaokunsun.github.io/MorphAny3D.github.io/.</p> + https://arxiv.org/abs/2601.00204 + Thu, 01 Jan 2026 04:42:59 +0000 - TimeBill: Time-Budgeted Inference for Large Language Models - https://arxiv.org/abs/2512.21859 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21859.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qi Fan, An Zou, Yehan Ma</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> Large Language Models (LLMs) are increasingly deployed in time-critical systems, such as robotics, autonomous driving, embodied intelligence, and industrial automation, where generating accurate responses within a given time budget is crucial for decision-making, control, or safety-critical tasks. However, the auto-regressive generation process of LLMs makes it challenging to model and estimate the end-to-end execution time. Furthermore, existing efficient inference methods based on a fixed key-value (KV) cache eviction ratio struggle to adapt to varying tasks with diverse time budgets, where an improper eviction ratio may lead to incomplete inference or a drop in response performance. In this paper, we propose TimeBill, a novel time-budgeted inference framework for LLMs that balances the inference efficiency and response performance. To be more specific, we propose a fine-grained response length predictor (RLP) and an execution time estimator (ETE) to accurately predict the end-to-end execution time of LLMs. Following this, we develop a time-budgeted efficient inference approach that adaptively adjusts the KV cache eviction ratio based on execution time prediction and the given time budget. Finally, through extensive experiments, we demonstrate the advantages of TimeBill in improving task completion rate and maintaining response performance under various overrun strategies.</p> - https://arxiv.org/abs/2512.21859 - Fri, 26 Dec 2025 04:49:35 +0000 + NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos + https://arxiv.org/abs/2601.00393 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00393.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuxue Yang, Lue Fan, Ziqi Shi, Junran Peng, Feng Wang, Zhaoxiang Zhang</p><p><b>Upvotes:</b> 109</p><p><b>Summary:</b> In this paper, we propose NeoVerse, a versatile 4D world model that is capable of 4D reconstruction, novel-trajectory video generation, and rich downstream applications. We first identify a common limitation of scalability in current 4D world modeling methods, caused either by expensive and specialized multi-view 4D data or by cumbersome training pre-processing. In contrast, our NeoVerse is built upon a core philosophy that makes the full pipeline scalable to diverse in-the-wild monocular videos. Specifically, NeoVerse features pose-free feed-forward 4D reconstruction, online monocular degradation pattern simulation, and other well-aligned techniques. These designs empower NeoVerse with versatility and generalization to various domains. Meanwhile, NeoVerse achieves state-of-the-art performance in standard reconstruction and generation benchmarks. Our project page is available at https://neoverse-4d.github.io</p> + https://arxiv.org/abs/2601.00393 + Thu, 01 Jan 2026 17:07:30 +0000 - SWE-RM: Execution-free Feedback For Software Engineering Agents - https://arxiv.org/abs/2512.21919 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21919.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> KaShun Shum, Binyuan Hui, Jiawei Chen, Lei Zhang, X. W., Jiaxi Yang, Yuzhen Huang, Junyang Lin, Junxian He</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Execution-based feedback like unit testing is widely used in the development of coding agents through test-time scaling (TTS) and reinforcement learning (RL). This paradigm requires scalable and reliable collection of unit test cases to provide accurate feedback, and the resulting feedback is often sparse and cannot effectively distinguish between trajectories that are both successful or both unsuccessful. In contrast, execution-free feedback from reward models can provide more fine-grained signals without depending on unit test cases. Despite this potential, execution-free feedback for realistic software engineering (SWE) agents remains underexplored. Aiming to develop versatile reward models that are effective across TTS and RL, however, we observe that two verifiers with nearly identical TTS performance can nevertheless yield very different results in RL. Intuitively, TTS primarily reflects the model's ability to select the best trajectory, but this ability does not necessarily generalize to RL. To address this limitation, we identify two additional aspects that are crucial for RL training: classification accuracy and calibration. We then conduct comprehensive controlled experiments to investigate how to train a robust reward model that performs well across these metrics. In particular, we analyze the impact of various factors such as training data scale, policy mixtures, and data source composition. Guided by these investigations, we introduce SWE-RM, an accurate and robust reward model adopting a mixture-of-experts architecture with 30B total parameters and 3B activated during inference. SWE-RM substantially improves SWE agents on both TTS and RL performance. For example, it increases the accuracy of Qwen3-Coder-Flash from 51.6% to 62.0%, and Qwen3-Coder-Max from 67.0% to 74.6% on SWE-Bench Verified using TTS, achieving new state-of-the-art performance among open-source models.</p> - https://arxiv.org/abs/2512.21919 - Fri, 26 Dec 2025 08:26:18 +0000 + Deep Delta Learning + https://arxiv.org/abs/2601.00417 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00417.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yifan Zhang, Yifeng Liu, Mengdi Wang, Quanquan Gu</p><p><b>Upvotes:</b> 29</p><p><b>Summary:</b> The efficacy of deep residual networks is fundamentally predicated on the identity shortcut connection. While this mechanism effectively mitigates the vanishing gradient problem, it imposes a strictly additive inductive bias on feature transformations, thereby limiting the network's capacity to model complex state transitions. In this paper, we introduce Deep Delta Learning (DDL), a novel architecture that generalizes the standard residual connection by modulating the identity shortcut with a learnable, data-dependent geometric transformation. This transformation, termed the Delta Operator, constitutes a rank-1 perturbation of the identity matrix, parameterized by a reflection direction vector k(X) and a gating scalar β(X). We provide a spectral analysis of this operator, demonstrating that the gate β(X) enables dynamic interpolation between identity mapping, orthogonal projection, and geometric reflection. Furthermore, we restructure the residual update as a synchronous rank-1 injection, where the gate acts as a dynamic step size governing both the erasure of old information and the writing of new features. This unification empowers the network to explicitly control the spectrum of its layer-wise transition operator, enabling the modeling of complex, non-monotonic dynamics while preserving the stable training characteristics of gated residual architectures.</p> + https://arxiv.org/abs/2601.00417 + Thu, 01 Jan 2026 18:11:38 +0000 - A 58-Addition, Rank-23 Scheme for General 3x3 Matrix Multiplication - https://arxiv.org/abs/2512.21980 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21980.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> A. I. Perminov</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> This paper presents a new state-of-the-art algorithm for exact 3times3 matrix multiplication over general non-commutative rings, achieving a rank-23 scheme with only 58 scalar additions. This improves the previous best additive complexity of 60 additions without a change of basis. The result was discovered through an automated search combining ternary-restricted flip-graph exploration with greedy intersection reduction for common subexpression elimination. The resulting scheme uses only coefficients from {-1, 0, 1}, ensuring both efficiency and portability across arbitrary fields. The total scalar operation count is reduced from 83 to 81.</p> - https://arxiv.org/abs/2512.21980 - Fri, 26 Dec 2025 10:58:54 +0000 + E-GRPO: High Entropy Steps Drive Effective Reinforcement Learning for Flow Models + https://arxiv.org/abs/2601.00423 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00423.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shengjun Zhang, Zhang Zhang, Chensheng Dai, Yueqi Duan</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Recent reinforcement learning has enhanced the flow matching models on human preference alignment. While stochastic sampling enables the exploration of denoising directions, existing methods which optimize over multiple denoising steps suffer from sparse and ambiguous reward signals. We observe that the high entropy steps enable more efficient and effective exploration while the low entropy steps result in undistinguished roll-outs. To this end, we propose E-GRPO, an entropy aware Group Relative Policy Optimization to increase the entropy of SDE sampling steps. Since the integration of stochastic differential equations suffer from ambiguous reward signals due to stochasticity from multiple steps, we specifically merge consecutive low entropy steps to formulate one high entropy step for SDE sampling, while applying ODE sampling on other steps. Building upon this, we introduce multi-step group normalized advantage, which computes group-relative advantages within samples sharing the same consolidated SDE denoising step. Experimental results on different reward settings have demonstrated the effectiveness of our methods.</p> + https://arxiv.org/abs/2601.00423 + Thu, 01 Jan 2026 18:27:32 +0000 - SmartSnap: Proactive Evidence Seeking for Self-Verifying Agents - https://arxiv.org/abs/2512.22322 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22322.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shaofei Cai, Yulei Qin, Haojia Lin, Zihan Xu, Gang Li, Yuchen Shi, Zongyi Li, Yong Mao, Siqi Cai, Xiaoyu Tan, Yitao Liang, Ke Li, Xing Sun</p><p><b>Upvotes:</b> 37</p><p><b>Summary:</b> Agentic reinforcement learning (RL) holds great promise for the development of autonomous agents under complex GUI tasks, but its scalability remains severely hampered by the verification of task completion. Existing task verification is treated as a passive, post-hoc process: a verifier (i.e., rule-based scoring script, reward or critic model, and LLM-as-a-Judge) analyzes the agent's entire interaction trajectory to determine if the agent succeeds. Such processing of verbose context that contains irrelevant, noisy history poses challenges to the verification protocols and therefore leads to prohibitive cost and low reliability. To overcome this bottleneck, we propose SmartSnap, a paradigm shift from this passive, post-hoc verification to proactive, in-situ self-verification by the agent itself. We introduce the Self-Verifying Agent, a new type of agent designed with dual missions: to not only complete a task but also to prove its accomplishment with curated snapshot evidences. Guided by our proposed 3C Principles (Completeness, Conciseness, and Creativity), the agent leverages its accessibility to the online environment to perform self-verification on a minimal, decisive set of snapshots. Such evidences are provided as the sole materials for a general LLM-as-a-Judge verifier to determine their validity and relevance. Experiments on mobile tasks across model families and scales demonstrate that our SmartSnap paradigm allows training LLM-driven agents in a scalable manner, bringing performance gains up to 26.08% and 16.66% respectively to 8B and 30B models. The synergizing between solution finding and evidence seeking facilitates the cultivation of efficient, self-verifying agents with competitive performance against DeepSeek V3.1 and Qwen3-235B-A22B.</p> - https://arxiv.org/abs/2512.22322 - Fri, 26 Dec 2025 14:51:39 +0000 + CPPO: Contrastive Perception for Vision Language Policy Optimization + https://arxiv.org/abs/2601.00501 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00501.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ahmad Rezaei, Mohsen Gholami, Saeed Ranjbar Alvar, Kevin Cannons, Mohammad Asiful Hossain, Zhou Weimin, Shunbo Zhou, Yong Zhang, Mohammad Akbari</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> We introduce CPPO, a Contrastive Perception Policy Optimization method for finetuning vision-language models (VLMs). While reinforcement learning (RL) has advanced reasoning in language models, extending it to multimodal reasoning requires improving both the perception and reasoning aspects. Prior works tackle this challenge mainly with explicit perception rewards, but disentangling perception tokens from reasoning tokens is difficult, requiring extra LLMs, ground-truth data, forced separation of perception from reasoning by policy model, or applying rewards indiscriminately to all output tokens. CPPO addresses this problem by detecting perception tokens via entropy shifts in the model outputs under perturbed input images. CPPO then extends the RL objective function with a Contrastive Perception Loss (CPL) that enforces consistency under information-preserving perturbations and sensitivity under information-removing ones. Experiments show that CPPO surpasses previous perception-rewarding methods, while avoiding extra models, making training more efficient and scalable.</p> + https://arxiv.org/abs/2601.00501 + Thu, 01 Jan 2026 22:48:26 +0000 - MAI-UI Technical Report: Real-World Centric Foundation GUI Agents - https://arxiv.org/abs/2512.22047 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22047.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hanzhang Zhou, Xu Zhang, Panrong Tong, Jianan Zhang, Liangyu Chen, Quyu Kong, Chenglin Cai, Chen Liu, Yue Wang, Jingren Zhou, Steven Hoi</p><p><b>Upvotes:</b> 26</p><p><b>Summary:</b> The development of GUI agents could revolutionize the next generation of human-computer interaction. Motivated by this vision, we present MAI-UI, a family of foundation GUI agents spanning the full spectrum of sizes, including 2B, 8B, 32B, and 235B-A22B variants. We identify four key challenges to realistic deployment: the lack of native agent-user interaction, the limits of UI-only operation, the absence of a practical deployment architecture, and brittleness in dynamic environments. MAI-UI addresses these issues with a unified methodology: a self-evolving data pipeline that expands the navigation data to include user interaction and MCP tool calls, a native device-cloud collaboration system routes execution by task state, and an online RL framework with advanced optimizations to scale parallel environments and context length. MAI-UI establishes new state-of-the-art across GUI grounding and mobile navigation. On grounding benchmarks, it reaches 73.5% on ScreenSpot-Pro, 91.3% on MMBench GUI L2, 70.9% on OSWorld-G, and 49.2% on UI-Vision, surpassing Gemini-3-Pro and Seed1.8 on ScreenSpot-Pro. On mobile GUI navigation, it sets a new SOTA of 76.7% on AndroidWorld, surpassing UI-Tars-2, Gemini-2.5-Pro and Seed1.8. On MobileWorld, MAI-UI obtains 41.7% success rate, significantly outperforming end-to-end GUI models and competitive with Gemini-3-Pro based agentic frameworks. Our online RL experiments show significant gains from scaling parallel environments from 32 to 512 (+5.2 points) and increasing environment step budget from 15 to 50 (+4.3 points). Finally, the native device-cloud collaboration system improves on-device performance by 33%, reduces cloud model calls by over 40%, and preserves user privacy.</p> - https://arxiv.org/abs/2512.22047 - Fri, 26 Dec 2025 14:51:52 +0000 + InfoSynth: Information-Guided Benchmark Synthesis for LLMs + https://arxiv.org/abs/2601.00575 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00575.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ishir Garg, Neel Kolhe, Xuandong Zhao, Dawn Song</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Large language models (LLMs) have demonstrated significant advancements in reasoning and code generation. However, efficiently creating new benchmarks to evaluate these capabilities remains a challenge. Traditional benchmark creation relies on manual human effort, a process that is both expensive and time-consuming. Furthermore, existing benchmarks often contaminate LLM training data, necessitating novel and diverse benchmarks to accurately assess their genuine capabilities. This work introduces InfoSynth, a novel framework for automatically generating and evaluating reasoning benchmarks guided by information-theoretic principles. We propose metrics based on KL-divergence and entropy to quantify benchmark novelty and diversity without relying on costly model evaluations. Building on this framework, we develop an end-to-end pipeline that synthesizes robust Python coding problems from seed datasets using genetic algorithms and iterative code feedback. Our method generates accurate test cases and solutions to new problems 97% of the time, and the synthesized benchmarks consistently exhibit higher novelty and diversity compared to their seed datasets. Moreover, our algorithm provides a method for controlling the novelty/diversity and difficulty of generated problems. InfoSynth offers a scalable, self-verifying pipeline for constructing high-quality, novel and diverse benchmarks for LLMs. Project Page: https://ishirgarg.github.io/infosynth_web/</p> + https://arxiv.org/abs/2601.00575 + Fri, 02 Jan 2026 05:26:27 +0000 - SpotEdit: Selective Region Editing in Diffusion Transformers - https://arxiv.org/abs/2512.22323 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22323.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhibin Qin, Zhenxiong Tan, Zeqing Wang, Songhua Liu, Xinchao Wang</p><p><b>Upvotes:</b> 37</p><p><b>Summary:</b> Diffusion Transformer models have significantly advanced image editing by encoding conditional images and integrating them into transformer layers. However, most edits involve modifying only small regions, while current methods uniformly process and denoise all tokens at every timestep, causing redundant computation and potentially degrading unchanged areas. This raises a fundamental question: Is it truly necessary to regenerate every region during editing? To address this, we propose SpotEdit, a training-free diffusion editing framework that selectively updates only the modified regions. SpotEdit comprises two key components: SpotSelector identifies stable regions via perceptual similarity and skips their computation by reusing conditional image features; SpotFusion adaptively blends these features with edited tokens through a dynamic fusion mechanism, preserving contextual coherence and editing quality. By reducing unnecessary computation and maintaining high fidelity in unmodified areas, SpotEdit achieves efficient and precise image editing.</p> - https://arxiv.org/abs/2512.22323 - Fri, 26 Dec 2025 14:59:41 +0000 + Avatar Forcing: Real-Time Interactive Head Avatar Generation for Natural Conversation + https://arxiv.org/abs/2601.00664 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00664.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Taekyung Ki, Sangwon Jang, Jaehyeong Jo, Jaehong Yoon, Sung Ju Hwang</p><p><b>Upvotes:</b> 50</p><p><b>Summary:</b> Talking head generation creates lifelike avatars from static portraits for virtual communication and content creation. However, current models do not yet convey the feeling of truly interactive communication, often generating one-way responses that lack emotional engagement. We identify two key challenges toward truly interactive avatars: generating motion in real-time under causal constraints and learning expressive, vibrant reactions without additional labeled data. To address these challenges, we propose Avatar Forcing, a new framework for interactive head avatar generation that models real-time user-avatar interactions through diffusion forcing. This design allows the avatar to process real-time multimodal inputs, including the user's audio and motion, with low latency for instant reactions to both verbal and non-verbal cues such as speech, nods, and laughter. Furthermore, we introduce a direct preference optimization method that leverages synthetic losing samples constructed by dropping user conditions, enabling label-free learning of expressive interaction. Experimental results demonstrate that our framework enables real-time interaction with low latency (approximately 500ms), achieving 6.8X speedup compared to the baseline, and produces reactive and expressive avatar motion, which is preferred over 80% against the baseline.</p> + https://arxiv.org/abs/2601.00664 + Fri, 02 Jan 2026 11:58:48 +0000 - Yume-1.5: A Text-Controlled Interactive World Generation Model - https://arxiv.org/abs/2512.22096 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22096.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiaofeng Mao, Zhen Li, Chuanhao Li, Xiaojie Xu, Kaining Ying, Tong He, Jiangmiao Pang, Yu Qiao, Kaipeng Zhang</p><p><b>Upvotes:</b> 56</p><p><b>Summary:</b> Recent approaches have demonstrated the promise of using diffusion models to generate interactive and explorable worlds. However, most of these methods face critical challenges such as excessively large parameter sizes, reliance on lengthy inference steps, and rapidly growing historical context, which severely limit real-time performance and lack text-controlled generation capabilities. To address these challenges, we propose \method, a novel framework designed to generate realistic, interactive, and continuous worlds from a single image or text prompt. \method achieves this through a carefully designed framework that supports keyboard-based exploration of the generated worlds. The framework comprises three core components: (1) a long-video generation framework integrating unified context compression with linear attention; (2) a real-time streaming acceleration strategy powered by bidirectional attention distillation and an enhanced text embedding scheme; (3) a text-controlled method for generating world events. We have provided the codebase in the supplementary material.</p> - https://arxiv.org/abs/2512.22096 - Fri, 26 Dec 2025 17:52:49 +0000 + Fast-weight Product Key Memory + https://arxiv.org/abs/2601.00671 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00671.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tianyu Zhao, Llion Jones</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Sequence modeling layers in modern language models typically face a trade-off between storage capacity and computational efficiency. While Softmax attention offers unbounded storage at prohibitive quadratic costs, linear variants provide efficiency but suffer from limited, fixed-size storage. We propose Fast-weight Product Key Memory (FwPKM), a novel architecture that resolves this tension by transforming the sparse Product Key Memory (PKM) from a static module into a dynamic, "fast-weight" episodic memory. Unlike PKM, FwPKM updates its parameters dynamically at both training and inference time via local chunk-level gradient descent, allowing the model to rapidly memorize and retrieve new key-value pairs from input sequences. Experiments reveal that FwPKM functions as an effective episodic memory that complements the semantic memory of standard modules, yielding significant perplexity reductions on long-context datasets. Notably, in Needle in a Haystack evaluations, FwPKM generalizes to 128K-token contexts despite being trained on only 4K-token sequences.</p> + https://arxiv.org/abs/2601.00671 + Fri, 02 Jan 2026 12:37:53 +0000 - Introducing TrGLUE and SentiTurca: A Comprehensive Benchmark for Turkish General Language Understanding and Sentiment Analysis - https://arxiv.org/abs/2512.22100 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22100.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Duygu Altinok</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Evaluating the performance of various model architectures, such as transformers, large language models (LLMs), and other NLP systems, requires comprehensive benchmarks that measure performance across multiple dimensions. Among these, the evaluation of natural language understanding (NLU) is particularly critical as it serves as a fundamental criterion for assessing model capabilities. Thus, it is essential to establish benchmarks that enable thorough evaluation and analysis of NLU abilities from diverse perspectives. While the GLUE benchmark has set a standard for evaluating English NLU, similar benchmarks have been developed for other languages, such as CLUE for Chinese, FLUE for French, and JGLUE for Japanese. However, no comparable benchmark currently exists for the Turkish language. To address this gap, we introduce TrGLUE, a comprehensive benchmark encompassing a variety of NLU tasks for Turkish. In addition, we present SentiTurca, a specialized benchmark for sentiment analysis. To support researchers, we also provide fine-tuning and evaluation code for transformer-based models, facilitating the effective use of these benchmarks. TrGLUE comprises Turkish-native corpora curated to mirror the domains and task formulations of GLUE-style evaluations, with labels obtained through a semi-automated pipeline that combines strong LLM-based annotation, cross-model agreement checks, and subsequent human validation. This design prioritizes linguistic naturalness, minimizes direct translation artifacts, and yields a scalable, reproducible workflow. With TrGLUE, our goal is to establish a robust evaluation framework for Turkish NLU, empower researchers with valuable resources, and provide insights into generating high-quality semi-automated datasets.</p> - https://arxiv.org/abs/2512.22100 - Fri, 26 Dec 2025 18:02:09 +0000 + The Reasoning-Creativity Trade-off: Toward Creativity-Driven Problem Solving + https://arxiv.org/abs/2601.00747 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00747.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Max Ruiz Luyten, Mihaela van der Schaar</p><p><b>Upvotes:</b> 18</p><p><b>Summary:</b> State-of-the-art large language model (LLM) pipelines rely on bootstrapped reasoning loops: sampling diverse chains of thought and reinforcing the highest-scoring ones, mainly optimizing correctness. We analyze how this design choice is sensitive to the collapse of the model's distribution over reasoning paths, slashing semantic entropy and undermining creative problem-solving. To analyze this failure, we introduce Distributional Creative Reasoning (DCR), a unified variational objective that casts training as gradient flow through probability measures on solution traces. STaR, GRPO, and DPO, as well as entropy bonuses, and other methods, all constitute special cases of the same loss. The framework delivers three core results: (i) the diversity decay theorem, describing how correctness-based objectives lead to distinct modes of diversity decay for STaR, GRPO, and DPO; (ii) designs that ensure convergence to a stable and diverse policy, effectively preventing collapse; and (iii) simple, actionable recipes to achieve this in practice. DCR thus offers the first principled recipe for LLMs that remain both correct and creative.</p> + https://arxiv.org/abs/2601.00747 + Fri, 02 Jan 2026 17:10:31 +0000 - ProEdit: Inversion-based Editing From Prompts Done Right - https://arxiv.org/abs/2512.22118 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22118.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhi Ouyang, Dian Zheng, Xiao-Ming Wu, Jian-Jian Jiang, Kun-Yu Lin, Jingke Meng, Wei-Shi Zheng</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Inversion-based visual editing provides an effective and training-free way to edit an image or a video based on user instructions. Existing methods typically inject source image information during the sampling process to maintain editing consistency. However, this sampling strategy overly relies on source information, which negatively affects the edits in the target image (e.g., failing to change the subject's atributes like pose, number, or color as instructed). In this work, we propose ProEdit to address this issue both in the attention and the latent aspects. In the attention aspect, we introduce KV-mix, which mixes KV features of the source and the target in the edited region, mitigating the influence of the source image on the editing region while maintaining background consistency. In the latent aspect, we propose Latents-Shift, which perturbs the edited region of the source latent, eliminating the influence of the inverted latent on the sampling. Extensive experiments on several image and video editing benchmarks demonstrate that our method achieves SOTA performance. In addition, our design is plug-and-play, which can be seamlessly integrated into existing inversion and editing methods, such as RF-Solver, FireFlow and UniEdit.</p> - https://arxiv.org/abs/2512.22118 - Fri, 26 Dec 2025 18:59:14 +0000 + AdaGaR: Adaptive Gabor Representation for Dynamic Scene Reconstruction + https://arxiv.org/abs/2601.00796 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.00796.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiewen Chan, Zhenjun Zhao, Yu-Lun Liu</p><p><b>Upvotes:</b> 29</p><p><b>Summary:</b> Reconstructing dynamic 3D scenes from monocular videos requires simultaneously capturing high-frequency appearance details and temporally continuous motion. Existing methods using single Gaussian primitives are limited by their low-pass filtering nature, while standard Gabor functions introduce energy instability. Moreover, lack of temporal continuity constraints often leads to motion artifacts during interpolation. We propose AdaGaR, a unified framework addressing both frequency adaptivity and temporal continuity in explicit dynamic scene modeling. We introduce Adaptive Gabor Representation, extending Gaussians through learnable frequency weights and adaptive energy compensation to balance detail capture and stability. For temporal continuity, we employ Cubic Hermite Splines with Temporal Curvature Regularization to ensure smooth motion evolution. An Adaptive Initialization mechanism combining depth estimation, point tracking, and foreground masks establishes stable point cloud distributions in early training. Experiments on Tap-Vid DAVIS demonstrate state-of-the-art performance (PSNR 35.49, SSIM 0.9433, LPIPS 0.0723) and strong generalization across frame interpolation, depth consistency, video editing, and stereo view synthesis. Project page: https://jiewenchan.github.io/AdaGaR/</p> + https://arxiv.org/abs/2601.00796 + Fri, 02 Jan 2026 18:59:55 +0000 - See Less, See Right: Bi-directional Perceptual Shaping For Multimodal Reasoning - https://arxiv.org/abs/2512.22120 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22120.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuoshuo Zhang, Yizhen Zhang, Jingjing Fu, Lei Song, Jiang Bian, Yujiu Yang, Rui Wang</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Large vision-language models (VLMs) often benefit from intermediate visual cues, either injected via external tools or generated as latent visual tokens during reasoning, but these mechanisms still overlook fine-grained visual evidence (e.g., polylines in charts), generalize poorly across domains, and incur high inference-time cost. In this paper, we propose Bi-directional Perceptual Shaping (BiPS), which transforms question-conditioned masked views into bidirectional where-to-look signals that shape perception during training. BiPS first applies a KL-consistency constraint between the original image and an evidence-preserving view that keeps only question-relevant regions, encouraging coarse but complete coverage of supporting pixels. It then applies a KL-separation constraint between the original and an evidence-ablated view where critical pixels are masked so the image no longer supports the original answer, discouraging text-only shortcuts (i.e., answering from text alone) and enforcing fine-grained visual reliance. Across eight benchmarks, BiPS boosts Qwen2.5-VL-7B by 8.2% on average and shows strong out-of-domain generalization to unseen datasets and image types.</p> - https://arxiv.org/abs/2512.22120 - Fri, 26 Dec 2025 18:59:47 +0000 + KV-Embedding: Training-free Text Embedding via Internal KV Re-routing in Decoder-only LLMs + https://arxiv.org/abs/2601.01046 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01046.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yixuan Tang, Yi Yang</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> While LLMs are powerful embedding backbones, their application in training-free settings faces two structural challenges: causal attention restricts early tokens from accessing subsequent context, and the next-token prediction objective biases representations toward generation rather than semantic compression. To address these limitations, we propose KV-Embedding, a framework that activates the latent representation power of frozen LLMs. Our method leverages the observation that the key-value (KV) states of the final token at each layer encode a compressed view of the sequence. By re-routing these states as a prepended prefix, we enable all tokens to access sequence-level context within a single forward pass. To ensure model-agnostic applicability, we introduce an automated layer selection strategy based on intrinsic dimensionality. Evaluations on MTEB across Qwen, Mistral, and Llama backbones show that KV-Embedding outperforms existing training-free baselines by up to 10%, while maintaining robust performance on sequences up to 4,096 tokens. These results demonstrate that internal state manipulation offers an efficient alternative to input modification, and we hope this work encourages further exploration of LLM internals for representation learning.</p> + https://arxiv.org/abs/2601.01046 + Sat, 03 Jan 2026 02:55:43 +0000 - VL-LN Bench: Towards Long-horizon Goal-oriented Navigation with Active Dialogs - https://arxiv.org/abs/2512.22342 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22342.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wensi Huang, Shaohao Zhu, Meng Wei, Jinming Xu, Xihui Liu, Hanqing Wang, Tai Wang, Feng Zhao, Jiangmiao Pang</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> In most existing embodied navigation tasks, instructions are well-defined and unambiguous, such as instruction following and object searching. Under this idealized setting, agents are required solely to produce effective navigation outputs conditioned on vision and language inputs. However, real-world navigation instructions are often vague and ambiguous, requiring the agent to resolve uncertainty and infer user intent through active dialog. To address this gap, we propose Interactive Instance Object Navigation (IION), a task that requires agents not only to generate navigation actions but also to produce language outputs via active dialog, thereby aligning more closely with practical settings. IION extends Instance Object Navigation (ION) by allowing agents to freely consult an oracle in natural language while navigating. Building on this task, we present the Vision Language-Language Navigation (VL-LN) benchmark, which provides a large-scale, automatically generated dataset and a comprehensive evaluation protocol for training and assessing dialog-enabled navigation models. VL-LN comprises over 41k long-horizon dialog-augmented trajectories for training and an automatic evaluation protocol with an oracle capable of responding to agent queries. Using this benchmark, we train a navigation model equipped with dialog capabilities and show that it achieves significant improvements over the baselines. Extensive experiments and analyses further demonstrate the effectiveness and reliability of VL-LN for advancing research on dialog-enabled embodied navigation. Code and dataset: https://0309hws.github.io/VL-LN.github.io/</p> - https://arxiv.org/abs/2512.22342 - Fri, 26 Dec 2025 19:00:12 +0000 + Digital Twin AI: Opportunities and Challenges from Large Language Models to World Models + https://arxiv.org/abs/2601.01321 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01321.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Rong Zhou, Dongping Chen, Zihan Jia, Yao Su, Yixin Liu, Yiwen Lu, Dongwei Shi, Yue Huang, Tianyang Xu, Yi Pan, Xinliang Li, Yohannes Abate, Qingyu Chen, Zhengzhong Tu, Yu Yang, Yu Zhang, Qingsong Wen, Gengchen Mai, Sunyang Fu, Jiachen Li, Xuyu Wang, Ziran Wang, Jing Huang, Tianming Liu, Yong Chen, Lichao Sun, Lifang He</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> Digital twins, as precise digital representations of physical systems, have evolved from passive simulation tools into intelligent and autonomous entities through the integration of artificial intelligence technologies. This paper presents a unified four-stage framework that systematically characterizes AI integration across the digital twin lifecycle, spanning modeling, mirroring, intervention, and autonomous management. By synthesizing existing technologies and practices, we distill a unified four-stage framework that systematically characterizes how AI methodologies are embedded across the digital twin lifecycle: (1) modeling the physical twin through physics-based and physics-informed AI approaches, (2) mirroring the physical system into a digital twin with real-time synchronization, (3) intervening in the physical twin through predictive modeling, anomaly detection, and optimization strategies, and (4) achieving autonomous management through large language models, foundation models, and intelligent agents. We analyze the synergy between physics-based modeling and data-driven learning, highlighting the shift from traditional numerical solvers to physics-informed and foundation models for physical systems. Furthermore, we examine how generative AI technologies, including large language models and generative world models, transform digital twins into proactive and self-improving cognitive systems capable of reasoning, communication, and creative scenario generation. Through a cross-domain review spanning eleven application domains, including healthcare, aerospace, smart manufacturing, robotics, and smart cities, we identify common challenges related to scalability, explainability, and trustworthiness, and outline directions for responsible AI-driven digital twin systems.</p> + https://arxiv.org/abs/2601.01321 + Sun, 04 Jan 2026 01:17:09 +0000 - Self-Evaluation Unlocks Any-Step Text-to-Image Generation - https://arxiv.org/abs/2512.22374 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22374.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xin Yu, Xiaojuan Qi, Zhengqi Li, Kai Zhang, Richard Zhang, Zhe Lin, Eli Shechtman, Tianyu Wang, Yotam Nitzan</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> We introduce the Self-Evaluating Model (Self-E), a novel, from-scratch training approach for text-to-image generation that supports any-step inference. Self-E learns from data similarly to a Flow Matching model, while simultaneously employing a novel self-evaluation mechanism: it evaluates its own generated samples using its current score estimates, effectively serving as a dynamic self-teacher. Unlike traditional diffusion or flow models, it does not rely solely on local supervision, which typically necessitates many inference steps. Unlike distillation-based approaches, it does not require a pretrained teacher. This combination of instantaneous local learning and self-driven global matching bridges the gap between the two paradigms, enabling the training of a high-quality text-to-image model from scratch that excels even at very low step counts. Extensive experiments on large-scale text-to-image benchmarks show that Self-E not only excels in few-step generation, but is also competitive with state-of-the-art Flow Matching models at 50 steps. We further find that its performance improves monotonically as inference steps increase, enabling both ultra-fast few-step generation and high-quality long-trajectory sampling within a single unified model. To our knowledge, Self-E is the first from-scratch, any-step text-to-image model, offering a unified framework for efficient and scalable generation.</p> - https://arxiv.org/abs/2512.22374 - Fri, 26 Dec 2025 20:42:11 +0000 + LEMAS: Large A 150K-Hour Large-scale Extensible Multilingual Audio Suite with Generative Speech Models + https://arxiv.org/abs/2601.04233 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04233.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhiyuan Zhao, Lijian Lin, Ye Zhu, Kai Xie, Yunfei Liu, Yu Li</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> We present the LEMAS-Dataset, which, to our knowledge, is currently the largest open-source multilingual speech corpus with word-level timestamps. Covering over 150,000 hours across 10 major languages, LEMAS-Dataset is constructed via a efficient data processing pipeline that ensures high-quality data and annotations. To validate the effectiveness of LEMAS-Dataset across diverse generative paradigms, we train two benchmark models with distinct architectures and task specializations on this dataset. LEMAS-TTS, built upon a non-autoregressive flow-matching framework, leverages the dataset's massive scale and linguistic diversity to achieve robust zero-shot multilingual synthesis. Our proposed accent-adversarial training and CTC loss mitigate cross-lingual accent issues, enhancing synthesis stability. Complementarily, LEMAS-Edit employs an autoregressive decoder-only architecture that formulates speech editing as a masked token infilling task. By exploiting precise word-level alignments to construct training masks and adopting adaptive decoding strategies, it achieves seamless, smooth-boundary speech editing with natural transitions. Experimental results demonstrate that models trained on LEMAS-Dataset deliver high-quality synthesis and editing performance, confirming the dataset's quality. We envision that this richly timestamp-annotated, fine-grained multilingual corpus will drive future advances in prompt-based speech generation systems.</p> + https://arxiv.org/abs/2601.04233 + Sun, 04 Jan 2026 03:43:52 +0000 - Monadic Context Engineering - https://arxiv.org/abs/2512.22431 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22431.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yifan Zhang, Mengdi Wang</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> The proliferation of Large Language Models (LLMs) has catalyzed a shift towards autonomous agents capable of complex reasoning and tool use. However, current agent architectures are frequently constructed using imperative, ad hoc patterns. This results in brittle systems plagued by difficulties in state management, error handling, and concurrency. This paper introduces Monadic Context Engineering (MCE), a novel architectural paradigm leveraging the algebraic structures of Functors, Applicative Functors, and Monads to provide a formal foundation for agent design. MCE treats agent workflows as computational contexts where cross-cutting concerns, such as state propagation, short-circuiting error handling, and asynchronous execution, are managed intrinsically by the algebraic properties of the abstraction. We demonstrate how Monads enable robust sequential composition, how Applicatives provide a principled structure for parallel execution, and crucially, how Monad Transformers allow for the systematic composition of these capabilities. This layered approach enables developers to construct complex, resilient, and efficient AI agents from simple, independently verifiable components. We further extend this framework to describe Meta-Agents, which leverage MCE for generative orchestration, dynamically creating and managing sub-agent workflows through metaprogramming. Project Page: https://github.com/yifanzhang-pro/monadic-context-engineering.</p> - https://arxiv.org/abs/2512.22431 - Sat, 27 Dec 2025 01:52:06 +0000 + DreamID-V:Bridging the Image-to-Video Gap for High-Fidelity Face Swapping via Diffusion Transformer + https://arxiv.org/abs/2601.01425 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01425.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xu Guo, Fulong Ye, Xinghui Li, Pengqi Tu, Pengze Zhang, Qichao Sun, Songtao Zhao, Xiangwang Hou, Qian He</p><p><b>Upvotes:</b> 48</p><p><b>Summary:</b> Video Face Swapping (VFS) requires seamlessly injecting a source identity into a target video while meticulously preserving the original pose, expression, lighting, background, and dynamic information. Existing methods struggle to maintain identity similarity and attribute preservation while preserving temporal consistency. To address the challenge, we propose a comprehensive framework to seamlessly transfer the superiority of Image Face Swapping (IFS) to the video domain. We first introduce a novel data pipeline SyncID-Pipe that pre-trains an Identity-Anchored Video Synthesizer and combines it with IFS models to construct bidirectional ID quadruplets for explicit supervision. Building upon paired data, we propose the first Diffusion Transformer-based framework DreamID-V, employing a core Modality-Aware Conditioning module to discriminatively inject multi-model conditions. Meanwhile, we propose a Synthetic-to-Real Curriculum mechanism and an Identity-Coherence Reinforcement Learning strategy to enhance visual realism and identity consistency under challenging scenarios. To address the issue of limited benchmarks, we introduce IDBench-V, a comprehensive benchmark encompassing diverse scenes. Extensive experiments demonstrate DreamID-V outperforms state-of-the-art methods and further exhibits exceptional versatility, which can be seamlessly adapted to various swap-related tasks.</p> + https://arxiv.org/abs/2601.01425 + Sun, 04 Jan 2026 08:07:11 +0000 - GraphLocator: Graph-guided Causal Reasoning for Issue Localization - https://arxiv.org/abs/2512.22469 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22469.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wei Liu, Chao Peng, Pengfei Gao, Aofan Liu, Wei Zhang, Haiyan Zhao, Zhi Jin</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> The issue localization task aims to identify the locations in a software repository that requires modification given a natural language issue description. This task is fundamental yet challenging in automated software engineering due to the semantic gap between issue description and source code implementation. This gap manifests as two mismatches:(1) symptom-to-cause mismatches, where descriptions do not explicitly reveal underlying root causes; (2) one-to-many mismatches, where a single issue corresponds to multiple interdependent code entities. To address these two mismatches, we propose GraphLocator, an approach that mitigates symptom-to-cause mismatches through causal structure discovering and resolves one-to-many mismatches via dynamic issue disentangling. The key artifact is the causal issue graph (CIG), in which vertices represent discovered sub-issues along with their associated code entities, and edges encode the causal dependencies between them. The workflow of GraphLocator consists of two phases: symptom vertices locating and dynamic CIG discovering; it first identifies symptom locations on the repository graph, then dynamically expands the CIG by iteratively reasoning over neighboring vertices. Experiments on three real-world datasets demonstrates the effectiveness of GraphLocator: (1) Compared with baselines, GraphLocator achieves more accurate localization with average improvements of +19.49% in function-level recall and +11.89% in precision. (2) GraphLocator outperforms baselines on both symptom-to-cause and one-to-many mismatch scenarios, achieving recall improvement of +16.44% and +19.18%, precision improvement of +7.78% and +13.23%, respectively. (3) The CIG generated by GraphLocator yields the highest relative improvement, resulting in a 28.74% increase in performance on downstream resolving task.</p> - https://arxiv.org/abs/2512.22469 - Sat, 27 Dec 2025 05:02:53 +0000 + SWE-Lego: Pushing the Limits of Supervised Fine-tuning for Software Issue Resolving + https://arxiv.org/abs/2601.01426 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01426.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chaofan Tao, Jierun Chen, Yuxin Jiang, Kaiqi Kou, Shaowei Wang, Ruoyu Wang, Xiaohui Li, Sidi Yang, Yiming Du, Jianbo Dai, Zhiming Mao, Xinyu Wang, Lifeng Shang, Haoli Bai</p><p><b>Upvotes:</b> 20</p><p><b>Summary:</b> We present SWE-Lego, a supervised fine-tuning (SFT) recipe designed to achieve state-ofthe-art performance in software engineering (SWE) issue resolving. In contrast to prevalent methods that rely on complex training paradigms (e.g., mid-training, SFT, reinforcement learning, and their combinations), we explore how to push the limits of a lightweight SFT-only approach for SWE tasks. SWE-Lego comprises three core building blocks, with key findings summarized as follows: 1) the SWE-Lego dataset, a collection of 32k highquality task instances and 18k validated trajectories, combining real and synthetic data to complement each other in both quality and quantity; 2) a refined SFT procedure with error masking and a difficulty-based curriculum, which demonstrably improves action quality and overall performance. Empirical results show that with these two building bricks alone,the SFT can push SWE-Lego models to state-of-the-art performance among open-source models of comparable size on SWE-bench Verified: SWE-Lego-Qwen3-8B reaches 42.2%, and SWE-Lego-Qwen3-32B attains 52.6%. 3) We further evaluate and improve test-time scaling (TTS) built upon the SFT foundation. Based on a well-trained verifier, SWE-Lego models can be significantly boosted--for example, 42.2% to 49.6% and 52.6% to 58.8% under TTS@16 for the 8B and 32B models, respectively.</p> + https://arxiv.org/abs/2601.01426 + Sun, 04 Jan 2026 08:07:27 +0000 - DreamOmni3: Scribble-based Editing and Generation - https://arxiv.org/abs/2512.22525 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22525.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bin Xia, Bohao Peng, Jiyang Liu, Sitong Wu, Jingyao Li, Junjia Huang, Xu Zhao, Yitong Wang, Ruihang Chu, Bei Yu, Jiaya Jia</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> Recently unified generation and editing models have achieved remarkable success with their impressive performance. These models rely mainly on text prompts for instruction-based editing and generation, but language often fails to capture users intended edit locations and fine-grained visual details. To this end, we propose two tasks: scribble-based editing and generation, that enables more flexible creation on graphical user interface (GUI) combining user textual, images, and freehand sketches. We introduce DreamOmni3, tackling two challenges: data creation and framework design. Our data synthesis pipeline includes two parts: scribble-based editing and generation. For scribble-based editing, we define four tasks: scribble and instruction-based editing, scribble and multimodal instruction-based editing, image fusion, and doodle editing. Based on DreamOmni2 dataset, we extract editable regions and overlay hand-drawn boxes, circles, doodles or cropped image to construct training data. For scribble-based generation, we define three tasks: scribble and instruction-based generation, scribble and multimodal instruction-based generation, and doodle generation, following similar data creation pipelines. For the framework, instead of using binary masks, which struggle with complex edits involving multiple scribbles, images, and instructions, we propose a joint input scheme that feeds both the original and scribbled source images into the model, using different colors to distinguish regions and simplify processing. By applying the same index and position encodings to both images, the model can precisely localize scribbled regions while maintaining accurate editing. Finally, we establish comprehensive benchmarks for these tasks to promote further research. Experimental results demonstrate that DreamOmni3 achieves outstanding performance, and models and code will be publicly released.</p> - https://arxiv.org/abs/2512.22525 - Sat, 27 Dec 2025 09:07:12 +0000 + MOSS Transcribe Diarize: Accurate Transcription with Speaker Diarization + https://arxiv.org/abs/2601.01554 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01554.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> MOSI. AI, Donghua Yu, Zhengyuan Lin, Chen Yang, Yiyang Zhang, Hanfu Chen, Jingqi Chen, Ke Chen, Liwei Fan, Yi Jiang, Jie Zhu, Muchen Li, Wenxuan Wang, Yang Wang, Zhe Xu, Yitian Gong, Yuqian Zhang, Wenbo Zhang, Zhaoye Fei, Qinyuan Cheng, Shimin Li, Xipeng Qiu</p><p><b>Upvotes:</b> 50</p><p><b>Summary:</b> Speaker-Attributed, Time-Stamped Transcription (SATS) aims to transcribe what is said and to precisely determine the timing of each speaker, which is particularly valuable for meeting transcription. Existing SATS systems rarely adopt an end-to-end formulation and are further constrained by limited context windows, weak long-range speaker memory, and the inability to output timestamps. To address these limitations, we present MOSS Transcribe Diarize, a unified multimodal large language model that jointly performs Speaker-Attributed, Time-Stamped Transcription in an end-to-end paradigm. Trained on extensive real wild data and equipped with a 128k context window for up to 90-minute inputs, MOSS Transcribe Diarize scales well and generalizes robustly. Across comprehensive evaluations, it outperforms state-of-the-art commercial systems on multiple public and in-house benchmarks.</p> + https://arxiv.org/abs/2601.01554 + Sun, 04 Jan 2026 15:01:10 +0000 - Geometry-Aware Optimization for Respiratory Sound Classification: Enhancing Sensitivity with SAM-Optimized Audio Spectrogram Transformers - https://arxiv.org/abs/2512.22564 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22564.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Atakan Işık, Selin Vulga Işık, Ahmet Feridun Işık, Mahşuk Taylan</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Respiratory sound classification is hindered by the limited size, high noise levels, and severe class imbalance of benchmark datasets like ICBHI 2017. While Transformer-based models offer powerful feature extraction capabilities, they are prone to overfitting and often converge to sharp minima in the loss landscape when trained on such constrained medical data. To address this, we introduce a framework that enhances the Audio Spectrogram Transformer (AST) using Sharpness-Aware Minimization (SAM). Instead of merely minimizing the training loss, our approach optimizes the geometry of the loss surface, guiding the model toward flatter minima that generalize better to unseen patients. We also implement a weighted sampling strategy to handle class imbalance effectively. Our method achieves a state-of-the-art score of 68.10% on the ICBHI 2017 dataset, outperforming existing CNN and hybrid baselines. More importantly, it reaches a sensitivity of 68.31%, a crucial improvement for reliable clinical screening. Further analysis using t-SNE and attention maps confirms that the model learns robust, discriminative features rather than memorizing background noise.</p> - https://arxiv.org/abs/2512.22564 - Sat, 27 Dec 2025 11:39:36 +0000 + OpenNovelty: An LLM-powered Agentic System for Verifiable Scholarly Novelty Assessment + https://arxiv.org/abs/2601.01576 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01576.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ming Zhang, Kexin Tan, Yueyuan Huang, Yujiong Shen, Chunchun Ma, Li Ju, Xinran Zhang, Yuhui Wang, Wenqing Jing, Jingyi Deng, Huayu Sha, Binze Hu, Jingqi Tong, Changhao Jiang, Yage Geng, Yuankai Ying, Yue Zhang, Zhangyue Yin, Zhiheng Xi, Shihan Dou, Tao Gui, Qi Zhang, Xuanjing Huang</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Evaluating novelty is critical yet challenging in peer review, as reviewers must assess submissions against a vast, rapidly evolving literature. This report presents OpenNovelty, an LLM-powered agentic system for transparent, evidence-based novelty analysis. The system operates through four phases: (1) extracting the core task and contribution claims to generate retrieval queries; (2) retrieving relevant prior work based on extracted queries via semantic search engine; (3) constructing a hierarchical taxonomy of core-task-related work and performing contribution-level full-text comparisons against each contribution; and (4) synthesizing all analyses into a structured novelty report with explicit citations and evidence snippets. Unlike naive LLM-based approaches, OpenNovelty grounds all assessments in retrieved real papers, ensuring verifiable judgments. We deploy our system on 500+ ICLR 2026 submissions with all reports publicly available on our website, and preliminary analysis suggests it can identify relevant prior work, including closely related papers that authors may overlook. OpenNovelty aims to empower the research community with a scalable tool that promotes fair, consistent, and evidence-backed peer review.</p> + https://arxiv.org/abs/2601.01576 + Sun, 04 Jan 2026 15:48:51 +0000 - Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbone - https://arxiv.org/abs/2512.22615 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22615.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiacheng Ye, Shansan Gong, Jiahui Gao, Junming Fan, Shuang Wu, Wei Bi, Haoli Bai, Lifeng Shang, Lingpeng Kong</p><p><b>Upvotes:</b> 43</p><p><b>Summary:</b> While autoregressive Large Vision-Language Models (VLMs) have achieved remarkable success, their sequential generation often limits their efficacy in complex visual planning and dynamic robotic control. In this work, we investigate the potential of constructing Vision-Language Models upon diffusion-based large language models (dLLMs) to overcome these limitations. We introduce Dream-VL, an open diffusion-based VLM (dVLM) that achieves state-of-the-art performance among previous dVLMs. Dream-VL is comparable to top-tier AR-based VLMs trained on open data on various benchmarks but exhibits superior potential when applied to visual planning tasks. Building upon Dream-VL, we introduce Dream-VLA, a dLLM-based Vision-Language-Action model (dVLA) developed through continuous pre-training on open robotic datasets. We demonstrate that the natively bidirectional nature of this diffusion backbone serves as a superior foundation for VLA tasks, inherently suited for action chunking and parallel generation, leading to significantly faster convergence in downstream fine-tuning. Dream-VLA achieves top-tier performance of 97.2% average success rate on LIBERO, 71.4% overall average on SimplerEnv-Bridge, and 60.5% overall average on SimplerEnv-Fractal, surpassing leading models such as π_0 and GR00T-N1. We also validate that dVLMs surpass AR baselines on downstream tasks across different training objectives. We release both Dream-VL and Dream-VLA to facilitate further research in the community.</p> - https://arxiv.org/abs/2512.22615 - Sat, 27 Dec 2025 14:46:24 +0000 + NitroGen: An Open Foundation Model for Generalist Gaming Agents + https://arxiv.org/abs/2601.02427 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02427.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Loïc Magne, Anas Awadalla, Guanzhi Wang, Yinzhen Xu, Joshua Belofsky, Fengyuan Hu, Joohwan Kim, Ludwig Schmidt, Georgia Gkioxari, Jan Kautz, Yisong Yue, Yejin Choi, Yuke Zhu, Linxi "Jim" Fan</p><p><b>Upvotes:</b> 35</p><p><b>Summary:</b> We introduce NitroGen, a vision-action foundation model for generalist gaming agents that is trained on 40,000 hours of gameplay videos across more than 1,000 games. We incorporate three key ingredients: 1) an internet-scale video-action dataset constructed by automatically extracting player actions from publicly available gameplay videos, 2) a multi-game benchmark environment that can measure cross-game generalization, and 3) a unified vision-action model trained with large-scale behavior cloning. NitroGen exhibits strong competence across diverse domains, including combat encounters in 3D action games, high-precision control in 2D platformers, and exploration in procedurally generated worlds. It transfers effectively to unseen games, achieving up to 52% relative improvement in task success rates over models trained from scratch. We release the dataset, evaluation suite, and model weights to advance research on generalist embodied agents.</p> + https://arxiv.org/abs/2601.02427 + Sun, 04 Jan 2026 16:24:50 +0000 - On the Role of Discreteness in Diffusion LLMs - https://arxiv.org/abs/2512.22630 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22630.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ziqi Jin, Bin Wang, Xiang Lin, Lidong Bing, Aixin Sun</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Diffusion models offer appealing properties for language generation, such as parallel decoding and iterative refinement, but the discrete and highly structured nature of text challenges the direct application of diffusion principles. In this paper, we revisit diffusion language modeling from the view of diffusion process and language modeling, and outline five properties that separate diffusion mechanics from language-specific requirements. We first categorize existing approaches into continuous diffusion in embedding space and discrete diffusion over tokens. We then show that each satisfies only part of the five essential properties and therefore reflects a structural trade-off. Through analyses of recent large diffusion language models, we identify two central issues: (i) uniform corruption does not respect how information is distributed across positions, and (ii) token-wise marginal training cannot capture multi-token dependencies during parallel decoding. These observations motivate diffusion processes that align more closely with the structure of text, and encourage future work toward more coherent diffusion language models.</p> - https://arxiv.org/abs/2512.22630 - Sat, 27 Dec 2025 16:03:08 +0000 + OpenRT: An Open-Source Red Teaming Framework for Multimodal LLMs + https://arxiv.org/abs/2601.01592 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01592.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xin Wang, Yunhao Chen, Juncheng Li, Yixu Wang, Yang Yao, Tianle Gu, Jie Li, Yan Teng, Xingjun Ma, Yingchun Wang, Xia Hu</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> The rapid integration of Multimodal Large Language Models (MLLMs) into critical applications is increasingly hindered by persistent safety vulnerabilities. However, existing red-teaming benchmarks are often fragmented, limited to single-turn text interactions, and lack the scalability required for systematic evaluation. To address this, we introduce OpenRT, a unified, modular, and high-throughput red-teaming framework designed for comprehensive MLLM safety evaluation. At its core, OpenRT architects a paradigm shift in automated red-teaming by introducing an adversarial kernel that enables modular separation across five critical dimensions: model integration, dataset management, attack strategies, judging methods, and evaluation metrics. By standardizing attack interfaces, it decouples adversarial logic from a high-throughput asynchronous runtime, enabling systematic scaling across diverse models. Our framework integrates 37 diverse attack methodologies, spanning white-box gradients, multi-modal perturbations, and sophisticated multi-agent evolutionary strategies. Through an extensive empirical study on 20 advanced models (including GPT-5.2, Claude 4.5, and Gemini 3 Pro), we expose critical safety gaps: even frontier models fail to generalize across attack paradigms, with leading models exhibiting average Attack Success Rates as high as 49.14%. Notably, our findings reveal that reasoning models do not inherently possess superior robustness against complex, multi-turn jailbreaks. By open-sourcing OpenRT, we provide a sustainable, extensible, and continuously maintained infrastructure that accelerates the development and standardization of AI safety.</p> + https://arxiv.org/abs/2601.01592 + Sun, 04 Jan 2026 16:41:33 +0000 - JavisGPT: A Unified Multi-modal LLM for Sounding-Video Comprehension and Generation - https://arxiv.org/abs/2512.22905 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22905.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Kai Liu, Jungang Li, Yuchong Sun, Shengqiong Wu, Jianzhang Gao, Daoan Zhang, Wei Zhang, Sheng Jin, Sicheng Yu, Geng Zhan, Jiayi Ji, Fan Zhou, Liang Zheng, Shuicheng Yan, Hao Fei, Tat-Seng Chua</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> This paper presents JavisGPT, the first unified multimodal large language model (MLLM) for Joint Audio-Video (JAV) comprehension and generation. JavisGPT adopts a concise encoder-LLM-decoder architecture, featuring a SyncFusion module for spatio-temporal audio-video fusion and synchrony-aware learnable queries to bridge a pretrained JAV-DiT generator. This design enables temporally coherent video-audio understanding and generation from multimodal instructions. We design an effective three-stage training pipeline consisting of multimodal pretraining, audio-video fine-tuning, and large-scale instruction-tuning, to progressively build multimodal comprehension and generation from existing vision-language models. To support this, we further construct JavisInst-Omni, a high-quality instruction dataset with over 200K GPT-4o-curated audio-video-text dialogues that span diverse and multi-level comprehension and generation scenarios. Extensive experiments on JAV comprehension and generation benchmarks show that JavisGPT outperforms existing MLLMs, particularly in complex and temporally synchronized settings.</p> - https://arxiv.org/abs/2512.22905 - Sun, 28 Dec 2025 12:25:43 +0000 + FFP-300K: Scaling First-Frame Propagation for Generalizable Video Editing + https://arxiv.org/abs/2601.01720 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01720.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xijie Huang, Chengming Xu, Donghao Luo, Xiaobin Hu, Peng Tang, Xu Peng, Jiangning Zhang, Chengjie Wang, Yanwei Fu</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> First-Frame Propagation (FFP) offers a promising paradigm for controllable video editing, but existing methods are hampered by a reliance on cumbersome run-time guidance. We identify the root cause of this limitation as the inadequacy of current training datasets, which are often too short, low-resolution, and lack the task diversity required to teach robust temporal priors. To address this foundational data gap, we first introduce FFP-300K, a new large-scale dataset comprising 300K high-fidelity video pairs at 720p resolution and 81 frames in length, constructed via a principled two-track pipeline for diverse local and global edits. Building on this dataset, we propose a novel framework designed for true guidance-free FFP that resolves the critical tension between maintaining first-frame appearance and preserving source video motion. Architecturally, we introduce Adaptive Spatio-Temporal RoPE (AST-RoPE), which dynamically remaps positional encodings to disentangle appearance and motion references. At the objective level, we employ a self-distillation strategy where an identity propagation task acts as a powerful regularizer, ensuring long-term temporal stability and preventing semantic drift. Comprehensive experiments on the EditVerseBench benchmark demonstrate that our method significantly outperforming existing academic and commercial models by receiving about 0.2 PickScore and 0.3 VLM score improvement against these competitors.</p> + https://arxiv.org/abs/2601.01720 + Mon, 05 Jan 2026 01:46:22 +0000 - Reverse Personalization - https://arxiv.org/abs/2512.22984 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.22984.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Han-Wei Kung, Tuomas Varanka, Nicu Sebe</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Recent text-to-image diffusion models have demonstrated remarkable generation of realistic facial images conditioned on textual prompts and human identities, enabling creating personalized facial imagery. However, existing prompt-based methods for removing or modifying identity-specific features rely either on the subject being well-represented in the pre-trained model or require model fine-tuning for specific identities. In this work, we analyze the identity generation process and introduce a reverse personalization framework for face anonymization. Our approach leverages conditional diffusion inversion, allowing direct manipulation of images without using text prompts. To generalize beyond subjects in the model's training data, we incorporate an identity-guided conditioning branch. Unlike prior anonymization methods, which lack control over facial attributes, our framework supports attribute-controllable anonymization. We demonstrate that our method achieves a state-of-the-art balance between identity removal, attribute preservation, and image quality. Source code and data are available at https://github.com/hanweikung/reverse-personalization .</p> - https://arxiv.org/abs/2512.22984 - Sun, 28 Dec 2025 16:06:55 +0000 + K-EXAONE Technical Report + https://arxiv.org/abs/2601.01739 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01739.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Eunbi Choi, Kibong Choi, Seokhee Hong, Junwon Hwang, Hyojin Jeon, Hyunjik Jo, Joonkee Kim, Seonghwan Kim, Soyeon Kim, Sunkyoung Kim, Yireun Kim, Yongil Kim, Haeju Lee, Jinsik Lee, Kyungmin Lee, Sangha Park, Heuiyeen Yeen, Hwan Chang, Stanley Jungkyu Choi, Yejin Choi, Jiwon Ham, Kijeong Jeon, Geunyeong Jeong, Gerrard Jeongwon Jo, Yonghwan Jo, Jiyeon Jung, Naeun Kang, Dohoon Kim, Euisoon Kim, Hayeon Kim, Hyosang Kim, Hyunseo Kim, Jieun Kim, Minu Kim, Myoungshin Kim, Unsol Kim, Youchul Kim, YoungJin Kim, Chaeeun Lee, Chaeyoon Lee, Changhun Lee, Dahm Lee, Edward Hwayoung Lee, Honglak Lee, Jinsang Lee, Jiyoung Lee, Sangeun Lee, Seungwon Lim, Solji Lim, Woohyung Lim, Chanwoo Moon, Jaewoo Park, Jinho Park, Yongmin Park, Hyerin Seo, Wooseok Seo, Yongwoo Song, Sejong Yang, Sihoon Yang, Chang En Yea, Sihyuk Yi, Chansik Yoon, Dongkeun Yoon, Sangyeon Yoon, Hyeongu Yun</p><p><b>Upvotes:</b> 78</p><p><b>Summary:</b> This technical report presents K-EXAONE, a large-scale multilingual language model developed by LG AI Research. K-EXAONE is built on a Mixture-of-Experts architecture with 236B total parameters, activating 23B parameters during inference. It supports a 256K-token context window and covers six languages: Korean, English, Spanish, German, Japanese, and Vietnamese. We evaluate K-EXAONE on a comprehensive benchmark suite spanning reasoning, agentic, general, Korean, and multilingual abilities. Across these evaluations, K-EXAONE demonstrates performance comparable to open-weight models of similar size. K-EXAONE, designed to advance AI for a better life, is positioned as a powerful proprietary AI foundation model for a wide range of industrial and research applications.</p> + https://arxiv.org/abs/2601.01739 + Mon, 05 Jan 2026 02:30:59 +0000 - Video-BrowseComp: Benchmarking Agentic Video Research on Open Web - https://arxiv.org/abs/2512.23044 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23044.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhengyang Liang, Yan Shu, Xiangrui Liu, Minghao Qin, Kaixin Liang, Paolo Rota, Nicu Sebe, Zheng Liu, Lizi Liao</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> The evolution of autonomous agents is redefining information seeking, transitioning from passive retrieval to proactive, open-ended web research. However, while textual and static multimodal agents have seen rapid progress, a significant modality gap remains in processing the web's most dynamic modality: video. Existing video benchmarks predominantly focus on passive perception, feeding curated clips to models without requiring external retrieval. They fail to evaluate agentic video research, which necessitates actively interrogating video timelines, cross-referencing dispersed evidence, and verifying claims against the open web. To bridge this gap, we present Video-BrowseComp, a challenging benchmark comprising 210 questions tailored for open-web agentic video reasoning. Unlike prior benchmarks, Video-BrowseComp enforces a mandatory dependency on temporal visual evidence, ensuring that answers cannot be derived solely through text search but require navigating video timelines to verify external claims. Our evaluation of state-of-the-art models reveals a critical bottleneck: even advanced search-augmented models like GPT-5.1 (w/ Search) achieve only 15.24\% accuracy. Our analysis reveals that these models largely rely on textual proxies, excelling in metadata-rich domains (e.g., TV shows with plot summaries) but collapsing in metadata-sparse, dynamic environments (e.g., sports, gameplay) where visual grounding is essential. As the first open-web video research benchmark, Video-BrowseComp advances the field beyond passive perception toward proactive video reasoning.</p> - https://arxiv.org/abs/2512.23044 - Sun, 28 Dec 2025 19:08:27 +0000 + COMPASS: A Framework for Evaluating Organization-Specific Policy Alignment in LLMs + https://arxiv.org/abs/2601.01836 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01836.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dasol Choi, DongGeon Lee, Brigitta Jesica Kartono, Helena Berndt, Taeyoun Kwon, Joonwon Jang, Haon Park, Hwanjo Yu, Minsuk Kahng</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> As large language models are deployed in high-stakes enterprise applications, from healthcare to finance, ensuring adherence to organization-specific policies has become essential. Yet existing safety evaluations focus exclusively on universal harms. We present COMPASS (Company/Organization Policy Alignment Assessment), the first systematic framework for evaluating whether LLMs comply with organizational allowlist and denylist policies. We apply COMPASS to eight diverse industry scenarios, generating and validating 5,920 queries that test both routine compliance and adversarial robustness through strategically designed edge cases. Evaluating seven state-of-the-art models, we uncover a fundamental asymmetry: models reliably handle legitimate requests (>95% accuracy) but catastrophically fail at enforcing prohibitions, refusing only 13-40% of adversarial denylist violations. These results demonstrate that current LLMs lack the robustness required for policy-critical deployments, establishing COMPASS as an essential evaluation framework for organizational AI safety.</p> + https://arxiv.org/abs/2601.01836 + Mon, 05 Jan 2026 06:57:45 +0000 - SurgWorld: Learning Surgical Robot Policies from Videos via World Modeling - https://arxiv.org/abs/2512.23162 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23162.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yufan He, Pengfei Guo, Mengya Xu, Zhaoshuo Li, Andriy Myronenko, Dillan Imans, Bingjie Liu, Dongren Yang, Mingxue Gu, Yongnan Ji, Yueming Jin, Ren Zhao, Baiyong Shen, Daguang Xu</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Data scarcity remains a fundamental barrier to achieving fully autonomous surgical robots. While large scale vision language action (VLA) models have shown impressive generalization in household and industrial manipulation by leveraging paired video action data from diverse domains, surgical robotics suffers from the paucity of datasets that include both visual observations and accurate robot kinematics. In contrast, vast corpora of surgical videos exist, but they lack corresponding action labels, preventing direct application of imitation learning or VLA training. In this work, we aim to alleviate this problem by learning policy models from SurgWorld, a world model designed for surgical physical AI. We curated the Surgical Action Text Alignment (SATA) dataset with detailed action description specifically for surgical robots. Then we built SurgeWorld based on the most advanced physical AI world model and SATA. It's able to generate diverse, generalizable and realistic surgery videos. We are also the first to use an inverse dynamics model to infer pseudokinematics from synthetic surgical videos, producing synthetic paired video action data. We demonstrate that a surgical VLA policy trained with these augmented data significantly outperforms models trained only on real demonstrations on a real surgical robot platform. Our approach offers a scalable path toward autonomous surgical skill acquisition by leveraging the abundance of unlabeled surgical video and generative world modeling, thus opening the door to generalizable and data efficient surgical robot policies.</p> - https://arxiv.org/abs/2512.23162 - Mon, 29 Dec 2025 03:03:00 +0000 + CogFlow: Bridging Perception and Reasoning through Knowledge Internalization for Visual Mathematical Problem Solving + https://arxiv.org/abs/2601.01874 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.01874.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuhang Chen, Yunqiu Xu, Junjie Xie, Aojun Lu, Tao Feng, Zeying Huang, Ning Zhang, Yi Sun, Yi Yang, Hangjie Yuan</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Despite significant progress, multimodal large language models continue to struggle with visual mathematical problem solving. Some recent works recognize that visual perception is a bottleneck in visual mathematical reasoning, but their solutions are limited to improving the extraction and interpretation of visual inputs. Notably, they all ignore the key issue of whether the extracted visual cues are faithfully integrated and properly utilized in subsequent reasoning. Motivated by this, we present CogFlow, a novel cognitive-inspired three-stage framework that incorporates a knowledge internalization stage, explicitly simulating the hierarchical flow of human reasoning: perceptionRightarrowinternalizationRightarrowreasoning. Inline with this hierarchical flow, we holistically enhance all its stages. We devise Synergistic Visual Rewards to boost perception capabilities in parametric and semantic spaces, jointly improving visual information extraction from symbols and diagrams. To guarantee faithful integration of extracted visual cues into subsequent reasoning, we introduce a Knowledge Internalization Reward model in the internalization stage, bridging perception and reasoning. Moreover, we design a Visual-Gated Policy Optimization algorithm to further enforce the reasoning is grounded with the visual knowledge, preventing models seeking shortcuts that appear coherent but are visually ungrounded reasoning chains. Moreover, we contribute a new dataset MathCog for model training, which contains samples with over 120K high-quality perception-reasoning aligned annotations. Comprehensive experiments and analysis on commonly used visual mathematical reasoning benchmarks validate the superiority of the proposed CogFlow.</p> + https://arxiv.org/abs/2601.01874 + Mon, 05 Jan 2026 08:02:18 +0000 - Evaluating Parameter Efficient Methods for RLVR - https://arxiv.org/abs/2512.23165 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23165.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qingyu Yin, Yulun Wu, Zhennan Shen, Sunbowen Li, Zhilin Wang, Yanshu Li, Chak Tou Leong, Jiale Kang, Jinjin Gu</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> We systematically evaluate Parameter-Efficient Fine-Tuning (PEFT) methods under the paradigm of Reinforcement Learning with Verifiable Rewards (RLVR). RLVR incentivizes language models to enhance their reasoning capabilities through verifiable feedback; however, while methods like LoRA are commonly used, the optimal PEFT architecture for RLVR remains unidentified. In this work, we conduct the first comprehensive evaluation of over 12 PEFT methodologies across the DeepSeek-R1-Distill families on mathematical reasoning benchmarks. Our empirical results challenge the default adoption of standard LoRA with three main findings. First, we demonstrate that structural variants, such as DoRA, AdaLoRA, and MiSS, consistently outperform LoRA. Second, we uncover a spectral collapse phenomenon in SVD-informed initialization strategies (e.g., PiSSA, MiLoRA), attributing their failure to a fundamental misalignment between principal-component updates and RL optimization. Furthermore, our ablations reveal that extreme parameter reduction (e.g., VeRA, Rank-1) severely bottlenecks reasoning capacity. We further conduct ablation studies and scaling experiments to validate our findings. This work provides a definitive guide for advocating for more exploration for parameter-efficient RL methods.</p> - https://arxiv.org/abs/2512.23165 - Mon, 29 Dec 2025 03:13:08 +0000 + WebGym: Scaling Training Environments for Visual Web Agents with Realistic Tasks + https://arxiv.org/abs/2601.02439 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02439.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hao Bai, Alexey Taymanov, Tong Zhang, Aviral Kumar, Spencer Whitehead</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> We present WebGym, the largest-to-date open-source environment for training realistic visual web agents. Real websites are non-stationary and diverse, making artificial or small-scale task sets insufficient for robust policy learning. WebGym contains nearly 300,000 tasks with rubric-based evaluations across diverse, real-world websites and difficulty levels. We train agents with a simple reinforcement learning (RL) recipe, which trains on the agent's own interaction traces (rollouts), using task rewards as feedback to guide learning. To enable scaling RL, we speed up sampling of trajectories in WebGym by developing a high-throughput asynchronous rollout system, designed specifically for web agents. Our system achieves a 4-5x rollout speedup compared to naive implementations. Second, we scale the task set breadth, depth, and size, which results in continued performance improvement. Fine-tuning a strong base vision-language model, Qwen-3-VL-8B-Instruct, on WebGym results in an improvement in success rate on an out-of-distribution test set from 26.2% to 42.9%, significantly outperforming agents based on proprietary models such as GPT-4o and GPT-5-Thinking that achieve 27.1% and 29.8%, respectively. This improvement is substantial because our test set consists only of tasks on websites never seen during training, unlike many other prior works on training visual web agents.</p> + https://arxiv.org/abs/2601.02439 + Mon, 05 Jan 2026 09:35:11 +0000 - Bridging Your Imagination with Audio-Video Generation via a Unified Director - https://arxiv.org/abs/2512.23222 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23222.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiaxu Zhang, Tianshu Hu, Yuan Zhang, Zenan Li, Linjie Luo, Guosheng Lin, Xin Chen</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Existing AI-driven video creation systems typically treat script drafting and key-shot design as two disjoint tasks: the former relies on large language models, while the latter depends on image generation models. We argue that these two tasks should be unified within a single framework, as logical reasoning and imaginative thinking are both fundamental qualities of a film director. In this work, we propose UniMAGE, a unified director model that bridges user prompts with well-structured scripts, thereby empowering non-experts to produce long-context, multi-shot films by leveraging existing audio-video generation models. To achieve this, we employ the Mixture-of-Transformers architecture that unifies text and image generation. To further enhance narrative logic and keyframe consistency, we introduce a ``first interleaving, then disentangling'' training paradigm. Specifically, we first perform Interleaved Concept Learning, which utilizes interleaved text-image data to foster the model's deeper understanding and imaginative interpretation of scripts. We then conduct Disentangled Expert Learning, which decouples script writing from keyframe generation, enabling greater flexibility and creativity in storytelling. Extensive experiments demonstrate that UniMAGE achieves state-of-the-art performance among open-source models, generating logically coherent video scripts and visually consistent keyframe images.</p> - https://arxiv.org/abs/2512.23222 - Mon, 29 Dec 2025 05:56:22 +0000 + MDAgent2: Large Language Model for Code Generation and Knowledge Q&A in Molecular Dynamics + https://arxiv.org/abs/2601.02075 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02075.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhuofan Shi, Hubao A, Yufei Shao, Mengyan Dai, Yadong Yu, Pan Xiang, Dongliang Huang, Hongxu An, Chunxiao Xin, Haiyang Shen, Zhenyu Wang, Yunshan Na, Gang Huang, Xiang Jing</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Molecular dynamics (MD) simulations are essential for understanding atomic-scale behaviors in materials science, yet writing LAMMPS scripts remains highly specialized and time-consuming tasks. Although LLMs show promise in code generation and domain-specific question answering, their performance in MD scenarios is limited by scarce domain data, the high deployment cost of state-of-the-art LLMs, and low code executability. Building upon our prior MDAgent, we present MDAgent2, the first end-to-end framework capable of performing both knowledge Q&A and code generation within the MD domain. We construct a domain-specific data-construction pipeline that yields three high-quality datasets spanning MD knowledge, question answering, and code generation. Based on these datasets, we adopt a three stage post-training strategy--continued pre-training (CPT), supervised fine-tuning (SFT), and reinforcement learning (RL)--to train two domain-adapted models, MD-Instruct and MD-Code. Furthermore, we introduce MD-GRPO, a closed-loop RL method that leverages simulation outcomes as reward signals and recycles low-reward trajectories for continual refinement. We further build MDAgent2-RUNTIME, a deployable multi-agent system that integrates code generation, execution, evaluation, and self-correction. Together with MD-EvalBench proposed in this work, the first benchmark for LAMMPS code generation and question answering, our models and system achieve performance surpassing several strong baselines.This work systematically demonstrates the adaptability and generalization capability of large language models in industrial simulation tasks, laying a methodological foundation for automatic code generation in AI for Science and industrial-scale simulations. URL: https://github.com/FredericVAN/PKU_MDAgent2</p> + https://arxiv.org/abs/2601.02075 + Mon, 05 Jan 2026 12:56:51 +0000 - KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta - https://arxiv.org/abs/2512.23236 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23236.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Gang Liao, Hongsen Qin, Ying Wang, Alicia Golden, Michael Kuchnik, Yavuz Yetim, Jia Jiunn Ang, Chunli Fu, Yihan He, Samuel Hsia, Zewei Jiang, Dianshi Li, Uladzimir Pashkevich, Varna Puvvada, Feng Shi, Matt Steiner, Ruichao Xiao, Nathan Yan, Xiayu Yu, Zhou Fang, Abdul Zainul-Abedin, Ketan Singh, Hongtao Yu, Wenyuan Chi, Barney Huang, Sean Zhang, Noah Weller, Zach Marine, Wyatt Cook, Carole-Jean Wu, Gaoxiang Liu</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Making deep learning recommendation model (DLRM) training and inference fast and efficient is important. However, this presents three key system challenges - model architecture diversity, kernel primitive diversity, and hardware generation and architecture heterogeneity. This paper presents KernelEvolve-an agentic kernel coding framework-to tackle heterogeneity at-scale for DLRM. KernelEvolve is designed to take kernel specifications as input and automate the process of kernel generation and optimization for recommendation model across heterogeneous hardware architectures. KernelEvolve does so by operating at multiple programming abstractions, from Triton and CuTe DSL to low-level hardware agnostic languages, spanning the full hardware-software optimization stack. The kernel optimization process is described as graph-based search with selection policy, universal operator, fitness function, and termination rule, dynamically adapts to runtime execution context through retrieval-augmented prompt synthesis. We designed, implemented, and deployed KernelEvolve to optimize a wide variety of production recommendation models across generations of NVIDIA and AMD GPUs, as well as Meta's AI accelerators. We validate KernelEvolve on the publicly-available KernelBench suite, achieving 100% pass rate on all 250 problems across three difficulty levels, and 160 PyTorch ATen operators across three heterogeneous hardware platforms, demonstrating 100% correctness. KernelEvolve reduces development time from weeks to hours and achieves substantial performance improvements over PyTorch baselines across diverse production use cases and for heterogeneous AI systems at-scale. Beyond performance efficiency improvements, KernelEvolve significantly mitigates the programmability barrier for new AI hardware by enabling automated kernel generation for in-house developed AI hardware.</p> - https://arxiv.org/abs/2512.23236 - Mon, 29 Dec 2025 06:31:55 +0000 + Entropy-Adaptive Fine-Tuning: Resolving Confident Conflicts to Mitigate Forgetting + https://arxiv.org/abs/2601.02151 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02151.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Muxi Diao, Lele Yang, Wuxuan Gong, Yutong Zhang, Zhonghao Yan, Yufei Han, Kongming Liang, Weiran Xu, Zhanyu Ma</p><p><b>Upvotes:</b> 87</p><p><b>Summary:</b> Supervised Fine-Tuning (SFT) is the standard paradigm for domain adaptation, yet it frequently incurs the cost of catastrophic forgetting. In sharp contrast, on-policy Reinforcement Learning (RL) effectively preserves general capabilities. We investigate this discrepancy and identify a fundamental distributional gap: while RL aligns with the model's internal belief, SFT forces the model to fit external supervision. This mismatch often manifests as "Confident Conflicts" tokens characterized by low probability but low entropy. In these instances, the model is highly confident in its own prediction but is forced to learn a divergent ground truth, triggering destructive gradient updates. To address this, we propose Entropy-Adaptive Fine-Tuning (EAFT). Unlike methods relying solely on prediction probability, EAFT utilizes token-level entropy as a gating mechanism to distinguish between epistemic uncertainty and knowledge conflict. This allows the model to learn from uncertain samples while suppressing gradients on conflicting data. Extensive experiments on Qwen and GLM series (ranging from 4B to 32B parameters) across mathematical, medical, and agentic domains confirm our hypothesis. EAFT consistently matches the downstream performance of standard SFT while significantly mitigating the degradation of general capabilities.</p> + https://arxiv.org/abs/2601.02151 + Mon, 05 Jan 2026 14:28:17 +0000 - YOLO-Master: MOE-Accelerated with Specialized Transformers for Enhanced Real-time Detection - https://arxiv.org/abs/2512.23273 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23273.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xu Lin, Jinlong Peng, Zhenye Gan, Jiawen Zhu, Jun Liu</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Existing Real-Time Object Detection (RTOD) methods commonly adopt YOLO-like architectures for their favorable trade-off between accuracy and speed. However, these models rely on static dense computation that applies uniform processing to all inputs, misallocating representational capacity and computational resources such as over-allocating on trivial scenes while under-serving complex ones. This mismatch results in both computational redundancy and suboptimal detection performance. To overcome this limitation, we propose YOLO-Master, a novel YOLO-like framework that introduces instance-conditional adaptive computation for RTOD. This is achieved through a Efficient Sparse Mixture-of-Experts (ES-MoE) block that dynamically allocates computational resources to each input according to its scene complexity. At its core, a lightweight dynamic routing network guides expert specialization during training through a diversity enhancing objective, encouraging complementary expertise among experts. Additionally, the routing network adaptively learns to activate only the most relevant experts, thereby improving detection performance while minimizing computational overhead during inference. Comprehensive experiments on five large-scale benchmarks demonstrate the superiority of YOLO-Master. On MS COCO, our model achieves 42.4% AP with 1.62ms latency, outperforming YOLOv13-N by +0.8% mAP and 17.8% faster inference. Notably, the gains are most pronounced on challenging dense scenes, while the model preserves efficiency on typical inputs and maintains real-time inference speed. Code will be available.</p> - https://arxiv.org/abs/2512.23273 - Mon, 29 Dec 2025 07:54:49 +0000 + Confidence Estimation for LLMs in Multi-turn Interactions + https://arxiv.org/abs/2601.02179 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02179.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Caiqi Zhang, Ruihan Yang, Xiaochen Zhu, Chengzu Li, Tiancheng Hu, Yijiang River Dong, Deqing Yang, Nigel Collier</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> While confidence estimation is a promising direction for mitigating hallucinations in Large Language Models (LLMs), current research dominantly focuses on single-turn settings. The dynamics of model confidence in multi-turn conversations, where context accumulates and ambiguity is progressively resolved, remain largely unexplored. Reliable confidence estimation in multi-turn settings is critical for many downstream applications, such as autonomous agents and human-in-the-loop systems. This work presents the first systematic study of confidence estimation in multi-turn interactions, establishing a formal evaluation framework grounded in two key desiderata: per-turn calibration and monotonicity of confidence as more information becomes available. To facilitate this, we introduce novel metrics, including a length-normalized Expected Calibration Error (InfoECE), and a new "Hinter-Guesser" paradigm for generating controlled evaluation datasets. Our experiments reveal that widely-used confidence techniques struggle with calibration and monotonicity in multi-turn dialogues. We propose P(Sufficient), a logit-based probe that achieves comparatively better performance, although the task remains far from solved. Our work provides a foundational methodology for developing more reliable and trustworthy conversational agents.</p> + https://arxiv.org/abs/2601.02179 + Mon, 05 Jan 2026 14:58:04 +0000 - AI Meets Brain: Memory Systems from Cognitive Neuroscience to Autonomous Agents - https://arxiv.org/abs/2512.23343 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23343.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiafeng Liang, Hao Li, Chang Li, Jiaqi Zhou, Shixin Jiang, Zekun Wang, Changkai Ji, Zhihao Zhu, Runxuan Liu, Tao Ren, Jinlan Fu, See-Kiong Ng, Xia Liang, Ming Liu, Bing Qin</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> Memory serves as the pivotal nexus bridging past and future, providing both humans and AI systems with invaluable concepts and experience to navigate complex tasks. Recent research on autonomous agents has increasingly focused on designing efficient memory workflows by drawing on cognitive neuroscience. However, constrained by interdisciplinary barriers, existing works struggle to assimilate the essence of human memory mechanisms. To bridge this gap, we systematically synthesizes interdisciplinary knowledge of memory, connecting insights from cognitive neuroscience with LLM-driven agents. Specifically, we first elucidate the definition and function of memory along a progressive trajectory from cognitive neuroscience through LLMs to agents. We then provide a comparative analysis of memory taxonomy, storage mechanisms, and the complete management lifecycle from both biological and artificial perspectives. Subsequently, we review the mainstream benchmarks for evaluating agent memory. Additionally, we explore memory security from dual perspectives of attack and defense. Finally, we envision future research directions, with a focus on multimodal memory systems and skill acquisition.</p> - https://arxiv.org/abs/2512.23343 - Mon, 29 Dec 2025 10:01:32 +0000 + NextFlow: Unified Sequential Modeling Activates Multimodal Understanding and Generation + https://arxiv.org/abs/2601.02204 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02204.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Huichao Zhang, Liao Qu, Yiheng Liu, Hang Chen, Yangyang Song, Yongsheng Dong, Shikun Sun, Xian Li, Xu Wang, Yi Jiang, Hu Ye, Bo Chen, Yiming Gao, Peng Liu, Akide Liu, Zhipeng Yang, Qili Deng, Linjie Xing, Jiyang Liu, Zhao Wang, Yang Zhou, Mingcong Liu, Yi Zhang, Qian He, Xiwei Hu, Zhongqi Qi, Jie Shao, Zhiye Fu, Shuai Wang, Fangmin Chen, Xuezhi Chai, Zhihua Wu, Yitong Wang, Zehuan Yuan, Daniel K. Du, Xinglong Wu</p><p><b>Upvotes:</b> 56</p><p><b>Summary:</b> We present NextFlow, a unified decoder-only autoregressive transformer trained on 6 trillion interleaved text-image discrete tokens. By leveraging a unified vision representation within a unified autoregressive architecture, NextFlow natively activates multimodal understanding and generation capabilities, unlocking abilities of image editing, interleaved content and video generation. Motivated by the distinct nature of modalities - where text is strictly sequential and images are inherently hierarchical - we retain next-token prediction for text but adopt next-scale prediction for visual generation. This departs from traditional raster-scan methods, enabling the generation of 1024x1024 images in just 5 seconds - orders of magnitude faster than comparable AR models. We address the instabilities of multi-scale generation through a robust training recipe. Furthermore, we introduce a prefix-tuning strategy for reinforcement learning. Experiments demonstrate that NextFlow achieves state-of-the-art performance among unified models and rivals specialized diffusion baselines in visual quality.</p> + https://arxiv.org/abs/2601.02204 + Mon, 05 Jan 2026 15:27:04 +0000 - A unified framework for detecting point and collective anomalies in operating system logs via collaborative transformers - https://arxiv.org/abs/2512.23380 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23380.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mohammad Nasirzadeh, Jafar Tahmoresnezhad, Parviz Rashidi-Khazaee</p><p><b>Upvotes:</b> 41</p><p><b>Summary:</b> Log anomaly detection is crucial for preserving the security of operating systems. Depending on the source of log data collection, various information is recorded in logs that can be considered log modalities. In light of this intuition, unimodal methods often struggle by ignoring the different modalities of log data. Meanwhile, multimodal methods fail to handle the interactions between these modalities. Applying multimodal sentiment analysis to log anomaly detection, we propose CoLog, a framework that collaboratively encodes logs utilizing various modalities. CoLog utilizes collaborative transformers and multi-head impressed attention to learn interactions among several modalities, ensuring comprehensive anomaly detection. To handle the heterogeneity caused by these interactions, CoLog incorporates a modality adaptation layer, which adapts the representations from different log modalities. This methodology enables CoLog to learn nuanced patterns and dependencies within the data, enhancing its anomaly detection capabilities. Extensive experiments demonstrate CoLog's superiority over existing state-of-the-art methods. Furthermore, in detecting both point and collective anomalies, CoLog achieves a mean precision of 99.63%, a mean recall of 99.59%, and a mean F1 score of 99.61% across seven benchmark datasets for log-based anomaly detection. The comprehensive detection capabilities of CoLog make it highly suitable for cybersecurity, system monitoring, and operational efficiency. CoLog represents a significant advancement in log anomaly detection, providing a sophisticated and effective solution to point and collective anomaly detection through a unified framework and a solution to the complex challenges automatic log data analysis poses. We also provide the implementation of CoLog at https://github.com/NasirzadehMoh/CoLog.</p> - https://arxiv.org/abs/2512.23380 - Mon, 29 Dec 2025 11:18:34 +0000 + VAR RL Done Right: Tackling Asynchronous Policy Conflicts in Visual Autoregressive Generation + https://arxiv.org/abs/2601.02256 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02256.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shikun Sun, Liao Qu, Huichao Zhang, Yiheng Liu, Yangyang Song, Xian Li, Xu Wang, Yi Jiang, Daniel K. Du, Xinglong Wu, Jia Jia</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> Visual generation is dominated by three paradigms: AutoRegressive (AR), diffusion, and Visual AutoRegressive (VAR) models. Unlike AR and diffusion, VARs operate on heterogeneous input structures across their generation steps, which creates severe asynchronous policy conflicts. This issue becomes particularly acute in reinforcement learning (RL) scenarios, leading to unstable training and suboptimal alignment. To resolve this, we propose a novel framework to enhance Group Relative Policy Optimization (GRPO) by explicitly managing these conflicts. Our method integrates three synergistic components: 1) a stabilizing intermediate reward to guide early-stage generation; 2) a dynamic time-step reweighting scheme for precise credit assignment; and 3) a novel mask propagation algorithm, derived from principles of Reward Feedback Learning (ReFL), designed to isolate optimization effects both spatially and temporally. Our approach demonstrates significant improvements in sample quality and objective alignment over the vanilla GRPO baseline, enabling robust and effective optimization for VAR models.</p> + https://arxiv.org/abs/2601.02256 + Mon, 05 Jan 2026 16:36:40 +0000 - Coupling Experts and Routers in Mixture-of-Experts via an Auxiliary Loss - https://arxiv.org/abs/2512.23447 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23447.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ang Lv, Jin Ma, Yiyuan Ma, Siyuan Qiao</p><p><b>Upvotes:</b> 88</p><p><b>Summary:</b> Mixture-of-Experts (MoE) models lack explicit constraints to ensure the router's decisions align well with the experts' capabilities, which ultimately limits model performance. To address this, we propose expert-router coupling (ERC) loss, a lightweight auxiliary loss that tightly couples the router's decisions with expert capabilities. Our approach treats each expert's router embedding as a proxy token for the tokens assigned to that expert, and feeds perturbed router embeddings through the experts to obtain internal activations. The ERC loss enforces two constraints on these activations: (1) Each expert must exhibit higher activation for its own proxy token than for the proxy tokens of any other expert. (2) Each proxy token must elicit stronger activation from its corresponding expert than from any other expert. These constraints jointly ensure that each router embedding faithfully represents its corresponding expert's capability, while each expert specializes in processing the tokens actually routed to it. The ERC loss is computationally efficient, operating only on n^2 activations, where n is the number of experts. This represents a fixed cost independent of batch size, unlike prior coupling methods that scale with the number of tokens (often millions per batch). Through pre-training MoE-LLMs ranging from 3B to 15B parameters and extensive analysis on trillions of tokens, we demonstrate the effectiveness of the ERC loss. Moreover, the ERC loss offers flexible control and quantitative tracking of expert specialization levels during training, providing valuable insights into MoEs.</p> - https://arxiv.org/abs/2512.23447 - Mon, 29 Dec 2025 13:03:18 +0000 + DiffProxy: Multi-View Human Mesh Recovery via Diffusion-Generated Dense Proxies + https://arxiv.org/abs/2601.02267 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02267.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Renke Wang, Zhenyu Zhang, Ying Tai, Jian Yang</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Human mesh recovery from multi-view images faces a fundamental challenge: real-world datasets contain imperfect ground-truth annotations that bias the models' training, while synthetic data with precise supervision suffers from domain gap. In this paper, we propose DiffProxy, a novel framework that generates multi-view consistent human proxies for mesh recovery. Central to DiffProxy is leveraging the diffusion-based generative priors to bridge the synthetic training and real-world generalization. Its key innovations include: (1) a multi-conditional mechanism for generating multi-view consistent, pixel-aligned human proxies; (2) a hand refinement module that incorporates flexible visual prompts to enhance local details; and (3) an uncertainty-aware test-time scaling method that increases robustness to challenging cases during optimization. These designs ensure that the mesh recovery process effectively benefits from the precise synthetic ground truth and generative advantages of the diffusion-based pipeline. Trained entirely on synthetic data, DiffProxy achieves state-of-the-art performance across five real-world benchmarks, demonstrating strong zero-shot generalization particularly on challenging scenarios with occlusions and partial views. Project page: https://wrk226.github.io/DiffProxy.html</p> + https://arxiv.org/abs/2601.02267 + Mon, 05 Jan 2026 16:51:45 +0000 - Act2Goal: From World Model To General Goal-conditioned Policy - https://arxiv.org/abs/2512.23541 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23541.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Pengfei Zhou, Liliang Chen, Shengcong Chen, Di Chen, Wenzhi Zhao, Rongjun Jin, Guanghui Ren, Jianlan Luo</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> Specifying robotic manipulation tasks in a manner that is both expressive and precise remains a central challenge. While visual goals provide a compact and unambiguous task specification, existing goal-conditioned policies often struggle with long-horizon manipulation due to their reliance on single-step action prediction without explicit modeling of task progress. We propose Act2Goal, a general goal-conditioned manipulation policy that integrates a goal-conditioned visual world model with multi-scale temporal control. Given a current observation and a target visual goal, the world model generates a plausible sequence of intermediate visual states that captures long-horizon structure. To translate this visual plan into robust execution, we introduce Multi-Scale Temporal Hashing (MSTH), which decomposes the imagined trajectory into dense proximal frames for fine-grained closed-loop control and sparse distal frames that anchor global task consistency. The policy couples these representations with motor control through end-to-end cross-attention, enabling coherent long-horizon behavior while remaining reactive to local disturbances. Act2Goal achieves strong zero-shot generalization to novel objects, spatial layouts, and environments. We further enable reward-free online adaptation through hindsight goal relabeling with LoRA-based finetuning, allowing rapid autonomous improvement without external supervision. Real-robot experiments demonstrate that Act2Goal improves success rates from 30% to 90% on challenging out-of-distribution tasks within minutes of autonomous interaction, validating that goal-conditioned world models with multi-scale temporal control provide structured guidance necessary for robust long-horizon manipulation. Project page: https://act2goal.github.io/</p> - https://arxiv.org/abs/2512.23541 - Mon, 29 Dec 2025 15:28:42 +0000 + InfiniteVGGT: Visual Geometry Grounded Transformer for Endless Streams + https://arxiv.org/abs/2601.02281 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02281.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuai Yuan, Yantai Yang, Xiaotian Yang, Xupeng Zhang, Zhonghao Zhao, Lingming Zhang, Zhipeng Zhang</p><p><b>Upvotes:</b> 29</p><p><b>Summary:</b> The grand vision of enabling persistent, large-scale 3D visual geometry understanding is shackled by the irreconcilable demands of scalability and long-term stability. While offline models like VGGT achieve inspiring geometry capability, their batch-based nature renders them irrelevant for live systems. Streaming architectures, though the intended solution for live operation, have proven inadequate. Existing methods either fail to support truly infinite-horizon inputs or suffer from catastrophic drift over long sequences. We shatter this long-standing dilemma with InfiniteVGGT, a causal visual geometry transformer that operationalizes the concept of a rolling memory through a bounded yet adaptive and perpetually expressive KV cache. Capitalizing on this, we devise a training-free, attention-agnostic pruning strategy that intelligently discards obsolete information, effectively ``rolling'' the memory forward with each new frame. Fully compatible with FlashAttention, InfiniteVGGT finally alleviates the compromise, enabling infinite-horizon streaming while outperforming existing streaming methods in long-term stability. The ultimate test for such a system is its performance over a truly infinite horizon, a capability that has been impossible to rigorously validate due to the lack of extremely long-term, continuous benchmarks. To address this critical gap, we introduce the Long3D benchmark, which, for the first time, enables a rigorous evaluation of continuous 3D geometry estimation on sequences about 10,000 frames. This provides the definitive evaluation platform for future research in long-term 3D geometry understanding. Code is available at: https://github.com/AutoLab-SAI-SJTU/InfiniteVGGT</p> + https://arxiv.org/abs/2601.02281 + Mon, 05 Jan 2026 17:11:00 +0000 - ProGuard: Towards Proactive Multimodal Safeguard - https://arxiv.org/abs/2512.23573 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23573.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shaohan Yu, Lijun Li, Chenyang Si, Lu Sheng, Jing Shao</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> The rapid evolution of generative models has led to a continuous emergence of multimodal safety risks, exposing the limitations of existing defense methods. To address these challenges, we propose ProGuard, a vision-language proactive guard that identifies and describes out-of-distribution (OOD) safety risks without the need for model adjustments required by traditional reactive approaches. We first construct a modality-balanced dataset of 87K samples, each annotated with both binary safety labels and risk categories under a hierarchical multimodal safety taxonomy, effectively mitigating modality bias and ensuring consistent moderation across text, image, and text-image inputs. Based on this dataset, we train our vision-language base model purely through reinforcement learning (RL) to achieve efficient and concise reasoning. To approximate proactive safety scenarios in a controlled setting, we further introduce an OOD safety category inference task and augment the RL objective with a synonym-bank-based similarity reward that encourages the model to generate concise descriptions for unseen unsafe categories. Experimental results show that ProGuard achieves performance comparable to closed-source large models on binary safety classification, substantially outperforms existing open-source guard models on unsafe content categorization. Most notably, ProGuard delivers a strong proactive moderation ability, improving OOD risk detection by 52.6% and OOD risk description by 64.8%.</p> - https://arxiv.org/abs/2512.23573 - Mon, 29 Dec 2025 16:13:23 +0000 + Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling + https://arxiv.org/abs/2601.02346 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02346.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Falcon LLM Team, Iheb Chaabane, Puneesh Khanna, Suhail Mohmad, Slim Frikha, Shi Hu, Abdalgader Abubaker, Reda Alami, Mikhail Lubinets, Mohamed El Amine Seddik, Hakim Hacid</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> This work introduces Falcon-H1R, a 7B-parameter reasoning-optimized model that establishes the feasibility of achieving competitive reasoning performance with small language models (SLMs). Falcon-H1R stands out for its parameter efficiency, consistently matching or outperforming SOTA reasoning models that are 2times to 7times larger across a variety of reasoning-intensive benchmarks. These results underscore the importance of careful data curation and targeted training strategies (via both efficient SFT and RL scaling) in delivering significant performance gains without increasing model size. Furthermore, Falcon-H1R advances the 3D limits of reasoning efficiency by combining faster inference (through its hybrid-parallel architecture design), token efficiency, and higher accuracy. This unique blend makes Falcon-H1R-7B a practical backbone for scaling advanced reasoning systems, particularly in scenarios requiring extensive chain-of-thoughts generation and parallel test-time scaling. Leveraging the recently introduced DeepConf approach, Falcon-H1R achieves state-of-the-art test-time scaling efficiency, offering substantial improvements in both accuracy and computational cost. As a result, Falcon-H1R demonstrates that compact models, through targeted model training and architectural choices, can deliver robust and scalable reasoning performance.</p> + https://arxiv.org/abs/2601.02346 + Mon, 05 Jan 2026 18:44:27 +0000 - LiveTalk: Real-Time Multimodal Interactive Video Diffusion via Improved On-Policy Distillation - https://arxiv.org/abs/2512.23576 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23576.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ethan Chern, Zhulin Hu, Bohao Tang, Jiadi Su, Steffi Chern, Zhijie Deng, Pengfei Liu</p><p><b>Upvotes:</b> 63</p><p><b>Summary:</b> Real-time video generation via diffusion is essential for building general-purpose multimodal interactive AI systems. However, the simultaneous denoising of all video frames with bidirectional attention via an iterative process in diffusion models prevents real-time interaction. While existing distillation methods can make the model autoregressive and reduce sampling steps to mitigate this, they focus primarily on text-to-video generation, leaving the human-AI interaction unnatural and less efficient. This paper targets real-time interactive video diffusion conditioned on a multimodal context, including text, image, and audio, to bridge the gap. Given the observation that the leading on-policy distillation approach Self Forcing encounters challenges (visual artifacts like flickering, black frames, and quality degradation) with multimodal conditioning, we investigate an improved distillation recipe with emphasis on the quality of condition inputs as well as the initialization and schedule for the on-policy optimization. On benchmarks for multimodal-conditioned (audio, image, and text) avatar video generation including HDTF, AVSpeech, and CelebV-HQ, our distilled model matches the visual quality of the full-step, bidirectional baselines of similar or larger size with 20x less inference cost and latency. Further, we integrate our model with audio language models and long-form video inference technique Anchor-Heavy Identity Sinks to build LiveTalk, a real-time multimodal interactive avatar system. System-level evaluation on our curated multi-turn interaction benchmark shows LiveTalk outperforms state-of-the-art models (Sora2, Veo3) in multi-turn video coherence and content quality, while reducing response latency from 1 to 2 minutes to real-time generation, enabling seamless human-AI multimodal interaction.</p> - https://arxiv.org/abs/2512.23576 - Mon, 29 Dec 2025 16:17:36 +0000 + Talk2Move: Reinforcement Learning for Text-Instructed Object-Level Geometric Transformation in Scenes + https://arxiv.org/abs/2601.02356 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02356.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jing Tan, Zhaoyang Zhang, Yantao Shen, Jiarui Cai, Shuo Yang, Jiajun Wu, Wei Xia, Zhuowen Tu, Stefano Soatto</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> We introduce Talk2Move, a reinforcement learning (RL) based diffusion framework for text-instructed spatial transformation of objects within scenes. Spatially manipulating objects in a scene through natural language poses a challenge for multimodal generation systems. While existing text-based manipulation methods can adjust appearance or style, they struggle to perform object-level geometric transformations-such as translating, rotating, or resizing objects-due to scarce paired supervision and pixel-level optimization limits. Talk2Move employs Group Relative Policy Optimization (GRPO) to explore geometric actions through diverse rollouts generated from input images and lightweight textual variations, removing the need for costly paired data. A spatial reward guided model aligns geometric transformations with linguistic description, while off-policy step evaluation and active step sampling improve learning efficiency by focusing on informative transformation stages. Furthermore, we design object-centric spatial rewards that evaluate displacement, rotation, and scaling behaviors directly, enabling interpretable and coherent transformations. Experiments on curated benchmarks demonstrate that Talk2Move achieves precise, consistent, and semantically faithful object transformations, outperforming existing text-guided editing approaches in both spatial accuracy and scene coherence.</p> + https://arxiv.org/abs/2601.02356 + Mon, 05 Jan 2026 18:55:32 +0000 - OmniAgent: Audio-Guided Active Perception Agent for Omnimodal Audio-Video Understanding - https://arxiv.org/abs/2512.23646 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23646.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Keda Tao, Wenjie Du, Bohan Yu, Weiqiang Wang, Jian Liu, Huan Wang</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> Omnimodal large language models have made significant strides in unifying audio and visual modalities; however, they often lack the fine-grained cross-modal understanding and have difficulty with multimodal alignment. To address these limitations, we introduce OmniAgent, a fully audio-guided active perception agent that dynamically orchestrates specialized tools to achieve more fine-grained audio-visual reasoning. Unlike previous works that rely on rigid, static workflows and dense frame-captioning, this paper demonstrates a paradigm shift from passive response generation to active multimodal inquiry. OmniAgent employs dynamic planning to autonomously orchestrate tool invocation on demand, strategically concentrating perceptual attention on task-relevant cues. Central to our approach is a novel coarse-to-fine audio-guided perception paradigm, which leverages audio cues to localize temporal events and guide subsequent reasoning. Extensive empirical evaluations on three audio-video understanding benchmarks demonstrate that OmniAgent achieves state-of-the-art performance, surpassing leading open-source and proprietary models by substantial margins of 10% - 20% accuracy.</p> - https://arxiv.org/abs/2512.23646 - Mon, 29 Dec 2025 17:59:05 +0000 + VINO: A Unified Visual Generator with Interleaved OmniModal Context + https://arxiv.org/abs/2601.02358 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02358.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Junyi Chen, Tong He, Zhoujie Fu, Pengfei Wan, Kun Gai, Weicai Ye</p><p><b>Upvotes:</b> 28</p><p><b>Summary:</b> We present VINO, a unified visual generator that performs image and video generation and editing within a single framework. Instead of relying on task-specific models or independent modules for each modality, VINO uses a shared diffusion backbone that conditions on text, images and videos, enabling a broad range of visual creation and editing tasks under one model. Specifically, VINO couples a vision-language model (VLM) with a Multimodal Diffusion Transformer (MMDiT), where multimodal inputs are encoded as interleaved conditioning tokens, and then used to guide the diffusion process. This design supports multi-reference grounding, long-form instruction following, and coherent identity preservation across static and dynamic content, while avoiding modality-specific architectural components. To train such a unified system, we introduce a multi-stage training pipeline that progressively expands a video generation base model into a unified, multi-task generator capable of both image and video input and output. Across diverse generation and editing benchmarks, VINO demonstrates strong visual quality, faithful instruction following, improved reference and attribute preservation, and more controllable multi-identity edits. Our results highlight a practical path toward scalable unified visual generation, and the promise of interleaved, in-context computation as a foundation for general-purpose visual creation.</p> + https://arxiv.org/abs/2601.02358 + Mon, 05 Jan 2026 18:56:34 +0000 - Nested Browser-Use Learning for Agentic Information Seeking - https://arxiv.org/abs/2512.23647 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23647.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Baixuan Li, Jialong Wu, Wenbiao Yin, Kuan Li, Zhongwang Zhang, Huifeng Yin, Zhengwei Tao, Liwen Zhang, Pengjun Xie, Jingren Zhou, Yong Jiang</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Information-seeking (IS) agents have achieved strong performance across a range of wide and deep search tasks, yet their tool use remains largely restricted to API-level snippet retrieval and URL-based page fetching, limiting access to the richer information available through real browsing. While full browser interaction could unlock deeper capabilities, its fine-grained control and verbose page content returns introduce substantial complexity for ReAct-style function-calling agents. To bridge this gap, we propose Nested Browser-Use Learning (NestBrowse), which introduces a minimal and complete browser-action framework that decouples interaction control from page exploration through a nested structure. This design simplifies agentic reasoning while enabling effective deep-web information acquisition. Empirical results on challenging deep IS benchmarks demonstrate that NestBrowse offers clear benefits in practice. Further in-depth analyses underscore its efficiency and flexibility.</p> - https://arxiv.org/abs/2512.23647 - Mon, 29 Dec 2025 17:59:14 +0000 + ExposeAnyone: Personalized Audio-to-Expression Diffusion Models Are Robust Zero-Shot Face Forgery Detectors + https://arxiv.org/abs/2601.02359 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02359.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Kaede Shiohara, Toshihiko Yamasaki, Vladislav Golyanik</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Detecting unknown deepfake manipulations remains one of the most challenging problems in face forgery detection. Current state-of-the-art approaches fail to generalize to unseen manipulations, as they primarily rely on supervised training with existing deepfakes or pseudo-fakes, which leads to overfitting to specific forgery patterns. In contrast, self-supervised methods offer greater potential for generalization, but existing work struggles to learn discriminative representations only from self-supervision. In this paper, we propose ExposeAnyone, a fully self-supervised approach based on a diffusion model that generates expression sequences from audio. The key idea is, once the model is personalized to specific subjects using reference sets, it can compute the identity distances between suspected videos and personalized subjects via diffusion reconstruction errors, enabling person-of-interest face forgery detection. Extensive experiments demonstrate that 1) our method outperforms the previous state-of-the-art method by 4.22 percentage points in the average AUC on DF-TIMIT, DFDCP, KoDF, and IDForge datasets, 2) our model is also capable of detecting Sora2-generated videos, where the previous approaches perform poorly, and 3) our method is highly robust to corruptions such as blur and compression, highlighting the applicability in real-world face forgery detection.</p> + https://arxiv.org/abs/2601.02359 + Mon, 05 Jan 2026 18:59:54 +0000 - End-to-End Test-Time Training for Long Context - https://arxiv.org/abs/2512.23675 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23675.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Arnuv Tandon, Karan Dalal, Xinhao Li, Daniel Koceja, Marcel Rød, Sam Buchanan, Xiaolong Wang, Jure Leskovec, Sanmi Koyejo, Tatsunori Hashimoto, Carlos Guestrin, Jed McCaleb, Yejin Choi, Yu Sun</p><p><b>Upvotes:</b> 14</p><p><b>Summary:</b> We formulate long-context language modeling as a problem in continual learning rather than architecture design. Under this formulation, we only use a standard architecture -- a Transformer with sliding-window attention. However, our model continues learning at test time via next-token prediction on the given context, compressing the context it reads into its weights. In addition, we improve the model's initialization for learning at test time via meta-learning at training time. Overall, our method, a form of Test-Time Training (TTT), is End-to-End (E2E) both at test time (via next-token prediction) and training time (via meta-learning), in contrast to previous forms. We conduct extensive experiments with a focus on scaling properties. In particular, for 3B models trained with 164B tokens, our method (TTT-E2E) scales with context length in the same way as Transformer with full attention, while others, such as Mamba 2 and Gated DeltaNet, do not. However, similar to RNNs, TTT-E2E has constant inference latency regardless of context length, making it 2.7 times faster than full attention for 128K context. Our code is publicly available.</p> - https://arxiv.org/abs/2512.23675 - Mon, 29 Dec 2025 18:30:14 +0000 + SimpleMem: Efficient Lifelong Memory for LLM Agents + https://arxiv.org/abs/2601.02553 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02553.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiaqi Liu, Yaofeng Su, Peng Xia, Siwei Han, Zeyu Zheng, Cihang Xie, Mingyu Ding, Huaxiu Yao</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> To support reliable long-term interaction in complex environments, LLM agents require memory systems that efficiently manage historical experiences. Existing approaches either retain full interaction histories via passive context extension, leading to substantial redundancy, or rely on iterative reasoning to filter noise, incurring high token costs. To address this challenge, we introduce SimpleMem, an efficient memory framework based on semantic lossless compression. We propose a three-stage pipeline designed to maximize information density and token utilization: (1) Semantic Structured Compression, which applies entropy-aware filtering to distill unstructured interactions into compact, multi-view indexed memory units; (2) Recursive Memory Consolidation, an asynchronous process that integrates related units into higher-level abstract representations to reduce redundancy; and (3) Adaptive Query-Aware Retrieval, which dynamically adjusts retrieval scope based on query complexity to construct precise context efficiently. Experiments on benchmark datasets show that our method consistently outperforms baseline approaches in accuracy, retrieval efficiency, and inference cost, achieving an average F1 improvement of 26.4% while reducing inference-time token consumption by up to 30-fold, demonstrating a superior balance between performance and efficiency. Code is available at https://github.com/aiming-lab/SimpleMem.</p> + https://arxiv.org/abs/2601.02553 + Mon, 05 Jan 2026 21:02:49 +0000 - Web World Models - https://arxiv.org/abs/2512.23676 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23676.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jichen Feng, Yifan Zhang, Chenggong Zhang, Yifu Lu, Shilong Liu, Mengdi Wang</p><p><b>Upvotes:</b> 20</p><p><b>Summary:</b> Language agents increasingly require persistent worlds in which they can act, remember, and learn. Existing approaches sit at two extremes: conventional web frameworks provide reliable but fixed contexts backed by databases, while fully generative world models aim for unlimited environments at the expense of controllability and practical engineering. In this work, we introduce the Web World Model (WWM), a middle ground where world state and ``physics'' are implemented in ordinary web code to ensure logical consistency, while large language models generate context, narratives, and high-level decisions on top of this structured latent state. We build a suite of WWMs on a realistic web stack, including an infinite travel atlas grounded in real geography, fictional galaxy explorers, web-scale encyclopedic and narrative worlds, and simulation- and game-like environments. Across these systems, we identify practical design principles for WWMs: separating code-defined rules from model-driven imagination, representing latent state as typed web interfaces, and utilizing deterministic generation to achieve unlimited but structured exploration. Our results suggest that web stacks themselves can serve as a scalable substrate for world models, enabling controllable yet open-ended environments. Project Page: https://github.com/Princeton-AI2-Lab/Web-World-Models.</p> - https://arxiv.org/abs/2512.23676 - Mon, 29 Dec 2025 18:31:45 +0000 + Learning User Preferences Through Interaction for Long-Term Collaboration + https://arxiv.org/abs/2601.02702 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02702.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuhaib Mehri, Priyanka Kargupta, Tal August, Dilek Hakkani-Tür</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> As conversational agents accumulate experience collaborating with users, adapting to user preferences is essential for fostering long-term relationships and improving collaboration quality over time. We introduce MultiSessionCollab, a benchmark that evaluates how well agents can learn user preferences and leverage them to improve collaboration quality throughout multiple sessions. To develop agents that succeed in this setting, we present long-term collaborative agents equipped with a memory that persists and refines user preference as interaction experience accumulates. Moreover, we demonstrate that learning signals can be derived from user simulator behavior in MultiSessionCollab to train agents to generate more comprehensive reflections and update their memory more effectively. Extensive experiments show that equipping agents with memory improves long-term collaboration, yielding higher task success rates, more efficient interactions, and reduced user effort. Finally, we conduct a human user study that demonstrates that memory helps improve user experience in real-world settings.</p> + https://arxiv.org/abs/2601.02702 + Tue, 06 Jan 2026 04:26:22 +0000 - Robo-Dopamine: General Process Reward Modeling for High-Precision Robotic Manipulation - https://arxiv.org/abs/2512.23703 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23703.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Huajie Tan, Sixiang Chen, Yijie Xu, Zixiao Wang, Yuheng Ji, Cheng Chi, Yaoxu Lyu, Zhongxia Zhao, Xiansheng Chen, Peterson Co, Shaoxuan Xie, Guocai Yao, Pengwei Wang, Zhongyuan Wang, Shanghang Zhang</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> The primary obstacle for applying reinforcement learning (RL) to real-world robotics is the design of effective reward functions. While recently learning-based Process Reward Models (PRMs) are a promising direction, they are often hindered by two fundamental limitations: their reward models lack step-aware understanding and rely on single-view perception, leading to unreliable assessments of fine-grained manipulation progress; and their reward shaping procedures are theoretically unsound, often inducing a semantic trap that misguides policy optimization. To address these, we introduce Dopamine-Reward, a novel reward modeling method for learning a general-purpose, step-aware process reward model from multi-view inputs. At its core is our General Reward Model (GRM), trained on a vast 3,400+ hour dataset, which leverages Step-wise Reward Discretization for structural understanding and Multi-Perspective Reward Fusion to overcome perceptual limitations. Building upon Dopamine-Reward, we propose Dopamine-RL, a robust policy learning framework that employs a theoretically-sound Policy-Invariant Reward Shaping method, which enables the agent to leverage dense rewards for efficient self-improvement without altering the optimal policy, thereby fundamentally avoiding the semantic trap. Extensive experiments across diverse simulated and real-world tasks validate our approach. GRM achieves state-of-the-art accuracy in reward assessment, and Dopamine-RL built on GRM significantly improves policy learning efficiency. For instance, after GRM is adapted to a new task in a one-shot manner from a single expert trajectory, the resulting reward model enables Dopamine-RL to improve the policy from near-zero to 95% success with only 150 online rollouts (approximately 1 hour of real robot interaction), while retaining strong generalization across tasks. Project website: https://robo-dopamine.github.io</p> - https://arxiv.org/abs/2512.23703 - Mon, 29 Dec 2025 18:57:44 +0000 + MiMo-V2-Flash Technical Report + https://arxiv.org/abs/2601.02780 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02780.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bangjun Xiao, Bingquan Xia, Bo Yang, Bofei Gao, Bowen Shen, Chen Zhang, Chenhong He, Chiheng Lou, Fuli Luo, Gang Wang, Gang Xie, Hailin Zhang, Hanglong Lv, Hanyu Li, Heyu Chen, Hongshen Xu, Houbin Zhang, Huaqiu Liu, Jiangshan Duo, Jianyu Wei, Jiebao Xiao, Jinhao Dong, Jun Shi, Junhao Hu, Kainan Bao, Kang Zhou, Lei Li, Liang Zhao, Linghao Zhang, Peidian Li, Qianli Chen, Shaohui Liu, Shihua Yu, Shijie Cao, Shimao Chen, Shouqiu Yu, Shuo Liu, Tianling Zhou, Weijiang Su, Weikun Wang, Wenhan Ma, Xiangwei Deng, Bohan Mao, Bowen Ye, Can Cai, Chenghua Wang, Chengxuan Zhu, Chong Ma, Chun Chen, Chunan Li, Dawei Zhu, Deshan Xiao, Dong Zhang, Duo Zhang, Fangyue Liu, Feiyu Yang, Fengyuan Shi, Guoan Wang, Hao Tian, Hao Wu, Heng Qu, Hongfei Yi, Hongxu An, Hongyi Guan, Xing Zhang, Yifan Song, Yihan Yan, Yihao Zhao, Yingchun Lai, Yizhao Gao, Yu Cheng, Yuanyuan Tian, Yudong Wang, Zhen Tang, Zhengju Tang, Zhengtao Wen, Zhichao Song, Zhixian Zheng, Zihan Jiang, Jian Wen, Jiarui Sun, Jiawei Li, Jinlong Xue, Jun Xia, Kai Fang, Menghang Zhu, Nuo Chen, Qian Tu, Qihao Zhang, Qiying Wang, Rang Li, Rui Ma, Shaolei Zhang, Shengfan Wang, Shicheng Li, Shuhao Gu, Shuhuai Ren, Sirui Deng, Tao Guo, Tianyang Lu, Weiji Zhuang, Weikang Zhang, Weimin Xiong, Wenshan Huang, Wenyu Yang, Xin Zhang, Xing Yong, Xu Wang, Xueyang Xie, Yilin Jiang, Yixin Yang, Yongzhe He, Yu Tu, Yuanliang Dong, Yuchen Liu, Yue Ma, Yue Yu, Yuxing Xiang, Zhaojun Huang, Zhenru Lin, Zhipeng Xu, Zhiyang Chen, Zhonghua Deng, Zihan Zhang, Zihao Yue</p><p><b>Upvotes:</b> 27</p><p><b>Summary:</b> We present MiMo-V2-Flash, a Mixture-of-Experts (MoE) model with 309B total parameters and 15B active parameters, designed for fast, strong reasoning and agentic capabilities. MiMo-V2-Flash adopts a hybrid attention architecture that interleaves Sliding Window Attention (SWA) with global attention, with a 128-token sliding window under a 5:1 hybrid ratio. The model is pre-trained on 27 trillion tokens with Multi-Token Prediction (MTP), employing a native 32k context length and subsequently extended to 256k. To efficiently scale post-training compute, MiMo-V2-Flash introduces a novel Multi-Teacher On-Policy Distillation (MOPD) paradigm. In this framework, domain-specialized teachers (e.g., trained via large-scale reinforcement learning) provide dense and token-level reward, enabling the student model to perfectly master teacher expertise. MiMo-V2-Flash rivals top-tier open-weight models such as DeepSeek-V3.2 and Kimi-K2, despite using only 1/2 and 1/3 of their total parameters, respectively. During inference, by repurposing MTP as a draft model for speculative decoding, MiMo-V2-Flash achieves up to 3.6 acceptance length and 2.6x decoding speedup with three MTP layers. We open-source both the model weights and the three-layer MTP weights to foster open research and community collaboration.</p> + https://arxiv.org/abs/2601.02780 + Tue, 06 Jan 2026 07:31:47 +0000 - Diffusion Knows Transparency: Repurposing Video Diffusion for Transparent Object Depth and Normal Estimation - https://arxiv.org/abs/2512.23705 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23705.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shaocong Xu, Songlin Wei, Qizhe Wei, Zheng Geng, Hong Li, Licheng Shen, Qianpu Sun, Shu Han, Bin Ma, Bohan Li, Chongjie Ye, Yuhang Zheng, Nan Wang, Saining Zhang, Hao Zhao</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> Transparent objects remain notoriously hard for perception systems: refraction, reflection and transmission break the assumptions behind stereo, ToF and purely discriminative monocular depth, causing holes and temporally unstable estimates. Our key observation is that modern video diffusion models already synthesize convincing transparent phenomena, suggesting they have internalized the optical rules. We build TransPhy3D, a synthetic video corpus of transparent/reflective scenes: 11k sequences rendered with Blender/Cycles. Scenes are assembled from a curated bank of category-rich static assets and shape-rich procedural assets paired with glass/plastic/metal materials. We render RGB + depth + normals with physically based ray tracing and OptiX denoising. Starting from a large video diffusion model, we learn a video-to-video translator for depth (and normals) via lightweight LoRA adapters. During training we concatenate RGB and (noisy) depth latents in the DiT backbone and co-train on TransPhy3D and existing frame-wise synthetic datasets, yielding temporally consistent predictions for arbitrary-length input videos. The resulting model, DKT, achieves zero-shot SOTA on real and synthetic video benchmarks involving transparency: ClearPose, DREDS (CatKnown/CatNovel), and TransPhy3D-Test. It improves accuracy and temporal consistency over strong image/video baselines, and a normal variant sets the best video normal estimation results on ClearPose. A compact 1.3B version runs at ~0.17 s/frame. Integrated into a grasping stack, DKT's depth boosts success rates across translucent, reflective and diffuse surfaces, outperforming prior estimators. Together, these results support a broader claim: "Diffusion knows transparency." Generative video priors can be repurposed, efficiently and label-free, into robust, temporally coherent perception for challenging real-world manipulation.</p> - https://arxiv.org/abs/2512.23705 - Mon, 29 Dec 2025 18:59:24 +0000 + DreamStyle: A Unified Framework for Video Stylization + https://arxiv.org/abs/2601.02785 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02785.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mengtian Li, Jinshu Chen, Songtao Zhao, Wanquan Feng, Pengqi Tu, Qian He</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> Video stylization, an important downstream task of video generation models, has not yet been thoroughly explored. Its input style conditions typically include text, style image, and stylized first frame. Each condition has a characteristic advantage: text is more flexible, style image provides a more accurate visual anchor, and stylized first frame makes long-video stylization feasible. However, existing methods are largely confined to a single type of style condition, which limits their scope of application. Additionally, their lack of high-quality datasets leads to style inconsistency and temporal flicker. To address these limitations, we introduce DreamStyle, a unified framework for video stylization, supporting (1) text-guided, (2) style-image-guided, and (3) first-frame-guided video stylization, accompanied by a well-designed data curation pipeline to acquire high-quality paired video data. DreamStyle is built on a vanilla Image-to-Video (I2V) model and trained using a Low-Rank Adaptation (LoRA) with token-specific up matrices that reduces the confusion among different condition tokens. Both qualitative and quantitative evaluations demonstrate that DreamStyle is competent in all three video stylization tasks, and outperforms the competitors in style consistency and video quality.</p> + https://arxiv.org/abs/2601.02785 + Tue, 06 Jan 2026 07:42:12 +0000 - Training AI Co-Scientists Using Rubric Rewards - https://arxiv.org/abs/2512.23707 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23707.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shashwat Goel, Rishi Hazra, Dulhan Jayalath, Timon Willi, Parag Jain, William F. Shen, Ilias Leontiadis, Francesco Barbieri, Yoram Bachrach, Jonas Geiping, Chenxi Whitehouse</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> AI co-scientists are emerging as a tool to assist human researchers in achieving their research goals. A crucial feature of these AI co-scientists is the ability to generate a research plan given a set of aims and constraints. The plan may be used by researchers for brainstorming, or may even be implemented after further refinement. However, language models currently struggle to generate research plans that follow all constraints and implicit requirements. In this work, we study how to leverage the vast corpus of existing research papers to train language models that generate better research plans. We build a scalable, diverse training corpus by automatically extracting research goals and goal-specific grading rubrics from papers across several domains. We then train models for research plan generation via reinforcement learning with self-grading. A frozen copy of the initial policy acts as the grader during training, with the rubrics creating a generator-verifier gap that enables improvements without external human supervision. To validate this approach, we conduct a study with human experts for machine learning research goals, spanning 225 hours. The experts prefer plans generated by our finetuned Qwen3-30B-A3B model over the initial model for 70% of research goals, and approve 84% of the automatically extracted goal-specific grading rubrics. To assess generality, we also extend our approach to research goals from medical papers, and new arXiv preprints, evaluating with a jury of frontier models. Our finetuning yields 12-22% relative improvements and significant cross-domain generalization, proving effective even in problem settings like medical research where execution feedback is infeasible. Together, these findings demonstrate the potential of a scalable, automated training recipe as a step towards improving general AI co-scientists.</p> - https://arxiv.org/abs/2512.23707 - Mon, 29 Dec 2025 18:59:33 +0000 + Mechanistic Interpretability of Large-Scale Counting in LLMs through a System-2 Strategy + https://arxiv.org/abs/2601.02989 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02989.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hosein Hasani, Mohammadali Banayeeanzade, Ali Nafisi, Sadegh Mohammadian, Fatemeh Askari, Mobin Bagherian, Amirmohammad Izadi, Mahdieh Soleymani Baghshah</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Large language models (LLMs), despite strong performance on complex mathematical problems, exhibit systematic limitations in counting tasks. This issue arises from architectural limits of transformers, where counting is performed across layers, leading to degraded precision for larger counting problems due to depth constraints. To address this limitation, we propose a simple test-time strategy inspired by System-2 cognitive processes that decomposes large counting tasks into smaller, independent sub-problems that the model can reliably solve. We evaluate this approach using observational and causal mediation analyses to understand the underlying mechanism of this System-2-like strategy. Our mechanistic analysis identifies key components: latent counts are computed and stored in the final item representations of each part, transferred to intermediate steps via dedicated attention heads, and aggregated in the final stage to produce the total count. Experimental results demonstrate that this strategy enables LLMs to surpass architectural limitations and achieve high accuracy on large-scale counting tasks. This work provides mechanistic insight into System-2 counting in LLMs and presents a generalizable approach for improving and understanding their reasoning behavior.</p> + https://arxiv.org/abs/2601.02989 + Tue, 06 Jan 2026 12:58:27 +0000 - Stream-DiffVSR: Low-Latency Streamable Video Super-Resolution via Auto-Regressive Diffusion - https://arxiv.org/abs/2512.23709 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23709.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hau-Shiang Shiu, Chin-Yang Lin, Zhixiang Wang, Chi-Wei Hsiao, Po-Fan Yu, Yu-Chih Chen, Yu-Lun Liu</p><p><b>Upvotes:</b> 44</p><p><b>Summary:</b> Diffusion-based video super-resolution (VSR) methods achieve strong perceptual quality but remain impractical for latency-sensitive settings due to reliance on future frames and expensive multi-step denoising. We propose Stream-DiffVSR, a causally conditioned diffusion framework for efficient online VSR. Operating strictly on past frames, it combines a four-step distilled denoiser for fast inference, an Auto-regressive Temporal Guidance (ARTG) module that injects motion-aligned cues during latent denoising, and a lightweight temporal-aware decoder with a Temporal Processor Module (TPM) that enhances detail and temporal coherence. Stream-DiffVSR processes 720p frames in 0.328 seconds on an RTX4090 GPU and significantly outperforms prior diffusion-based methods. Compared with the online SOTA TMP, it boosts perceptual quality (LPIPS +0.095) while reducing latency by over 130x. Stream-DiffVSR achieves the lowest latency reported for diffusion-based VSR, reducing initial delay from over 4600 seconds to 0.328 seconds, thereby making it the first diffusion VSR method suitable for low-latency online deployment. Project page: https://jamichss.github.io/stream-diffvsr-project-page/</p> - https://arxiv.org/abs/2512.23709 - Mon, 29 Dec 2025 18:59:57 +0000 + Large Reasoning Models Are (Not Yet) Multilingual Latent Reasoners + https://arxiv.org/abs/2601.02996 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.02996.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yihong Liu, Raoyuan Zhao, Hinrich Schütze, Michael A. Hedderich</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Large reasoning models (LRMs) achieve strong performance on mathematical reasoning tasks, often attributed to their capability to generate explicit chain-of-thought (CoT) explanations. However, recent work shows that LRMs often arrive at the correct answer before completing these textual reasoning steps, indicating the presence of latent reasoning -- internal, non-verbal computation encoded in hidden states. While this phenomenon has been explored in English, its multilingual behavior remains largely unknown. In this paper, we conduct a systematic investigation of multilingual latent reasoning in LRMs across 11 languages. Using a truncation-based strategy, we examine how the correct answer emerges as the model is given only partial reasoning traces, allowing us to measure stepwise latent prediction formation. Our results reveal clear evidence of multilingual latent reasoning, though unevenly: strong in resource-rich languages, weaker in low-resource ones, and broadly less observable on harder benchmarks. To understand whether these differences reflect distinct internal mechanisms, we further perform representational analyses. Despite surface-level disparities, we find that the internal evolution of predictions is highly consistent across languages and broadly aligns with English -- a pattern suggesting an English-centered latent reasoning pathway.</p> + https://arxiv.org/abs/2601.02996 + Tue, 06 Jan 2026 13:20:17 +0000 - Pretraining Frame Preservation in Autoregressive Video Memory Compression - https://arxiv.org/abs/2512.23851 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23851.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Lvmin Zhang, Shengqu Cai, Muyang Li, Chong Zeng, Beijia Lu, Anyi Rao, Song Han, Gordon Wetzstein, Maneesh Agrawala</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> We present PFP, a neural network structure to compress long videos into short contexts, with an explicit pretraining objective to preserve the high-frequency details of single frames at arbitrary temporal positions. The baseline model can compress a 20-second video into a context at about 5k length, where random frames can be retrieved with perceptually preserved appearances. Such pretrained models can be directly fine-tuned as memory encoders for autoregressive video models, enabling long history memory with low context cost and relatively low fidelity loss. We evaluate the framework with ablative settings and discuss the trade-offs of possible neural architecture designs.</p> - https://arxiv.org/abs/2512.23851 - Mon, 29 Dec 2025 20:29:21 +0000 + Why LLMs Aren't Scientists Yet: Lessons from Four Autonomous Research Attempts + https://arxiv.org/abs/2601.03315 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03315.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dhruv Trehan, Paras Chopra</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> We report a case study of four end-to-end attempts to autonomously generate ML research papers using a pipeline of six LLM agents mapped to stages of the scientific workflow. Of these four, three attempts failed during implementation or evaluation. One completed the pipeline and was accepted to Agents4Science 2025, an experimental inaugural venue that required AI systems as first authors, passing both human and multi-AI review. From these attempts, we document six recurring failure modes: bias toward training data defaults, implementation drift under execution pressure, memory and context degradation across long-horizon tasks, overexcitement that declares success despite obvious failures, insufficient domain intelligence, and weak scientific taste in experimental design. We conclude by discussing four design principles for more robust AI-scientist systems, implications for autonomous scientific discovery, and we release all prompts, artifacts, and outputs at https://github.com/Lossfunk/ai-scientist-artefacts-v1</p> + https://arxiv.org/abs/2601.03315 + Tue, 06 Jan 2026 13:20:54 +0000 - Improving Multi-step RAG with Hypergraph-based Memory for Long-Context Complex Relational Modeling - https://arxiv.org/abs/2512.23959 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23959.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chulun Zhou, Chunkang Zhang, Guoxin Yu, Fandong Meng, Jie Zhou, Wai Lam, Mo Yu</p><p><b>Upvotes:</b> 83</p><p><b>Summary:</b> Multi-step retrieval-augmented generation (RAG) has become a widely adopted strategy for enhancing large language models (LLMs) on tasks that demand global comprehension and intensive reasoning. Many RAG systems incorporate a working memory module to consolidate retrieved information. However, existing memory designs function primarily as passive storage that accumulates isolated facts for the purpose of condensing the lengthy inputs and generating new sub-queries through deduction. This static nature overlooks the crucial high-order correlations among primitive facts, the compositions of which can often provide stronger guidance for subsequent steps. Therefore, their representational strength and impact on multi-step reasoning and knowledge evolution are limited, resulting in fragmented reasoning and weak global sense-making capacity in extended contexts. We introduce HGMem, a hypergraph-based memory mechanism that extends the concept of memory beyond simple storage into a dynamic, expressive structure for complex reasoning and global understanding. In our approach, memory is represented as a hypergraph whose hyperedges correspond to distinct memory units, enabling the progressive formation of higher-order interactions within memory. This mechanism connects facts and thoughts around the focal problem, evolving into an integrated and situated knowledge structure that provides strong propositions for deeper reasoning in subsequent steps. We evaluate HGMem on several challenging datasets designed for global sense-making. Extensive experiments and in-depth analyses show that our method consistently improves multi-step RAG and substantially outperforms strong baseline systems across diverse tasks.</p> - https://arxiv.org/abs/2512.23959 - Tue, 30 Dec 2025 03:13:10 +0000 + SOP: A Scalable Online Post-Training System for Vision-Language-Action Models + https://arxiv.org/abs/2601.03044 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03044.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingjie Pan, Siyuan Feng, Qinglin Zhang, Xinchen Li, Jianheng Song, Chendi Qu, Yi Wang, Chuankang Li, Ziyu Xiong, Zhi Chen, Yi Liu, Jianlan Luo</p><p><b>Upvotes:</b> 26</p><p><b>Summary:</b> Vision-language-action (VLA) models achieve strong generalization through large-scale pre-training, but real-world deployment requires expert-level task proficiency in addition to broad generality. Existing post-training approaches for VLA models are typically offline, single-robot, or task-specific, limiting effective on-policy adaptation and scalable learning from real-world interaction. We introduce a Scalable Online Post-training (SOP) system that enables online, distributed, multi-task post-training of generalist VLA models directly in the physical world. SOP tightly couples execution and learning through a closed-loop architecture in which a fleet of robots continuously streams on-policy experience and human intervention signals to a centralized cloud learner, and asynchronously receives updated policies. This design supports prompt on-policy correction, scales experience collection through parallel deployment, and preserves generality during adaptation. SOP is agnostic to the choice of post-training algorithm; we instantiate it with both interactive imitation learning (HG-DAgger) and reinforcement learning (RECAP). Across a range of real-world manipulation tasks including cloth folding, box assembly, and grocery restocking, we show that SOP substantially improves the performance of large pretrained VLA models while maintaining a single shared policy across tasks. Effective post-training can be achieved within hours of real-world interaction, and performance scales near-linearly with the number of robots in the fleet. These results suggest that tightly coupling online learning with fleet-scale deployment is instrumental to enabling efficient, reliable, and scalable post-training of generalist robot policies in the physical world.</p> + https://arxiv.org/abs/2601.03044 + Tue, 06 Jan 2026 14:25:11 +0000 - Fantastic Reasoning Behaviors and Where to Find Them: Unsupervised Discovery of the Reasoning Process - https://arxiv.org/abs/2512.23988 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.23988.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhenyu Zhang, Shujian Zhang, John Lambert, Wenxuan Zhou, Zhangyang Wang, Mingqing Chen, Andrew Hard, Rajiv Mathews, Lun Wang</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Despite the growing reasoning capabilities of recent large language models (LLMs), their internal mechanisms during the reasoning process remain underexplored. Prior approaches often rely on human-defined concepts (e.g., overthinking, reflection) at the word level to analyze reasoning in a supervised manner. However, such methods are limited, as it is infeasible to capture the full spectrum of potential reasoning behaviors, many of which are difficult to define in token space. In this work, we propose an unsupervised framework (namely, RISE: Reasoning behavior Interpretability via Sparse auto-Encoder) for discovering reasoning vectors, which we define as directions in the activation space that encode distinct reasoning behaviors. By segmenting chain-of-thought traces into sentence-level 'steps' and training sparse auto-encoders (SAEs) on step-level activations, we uncover disentangled features corresponding to interpretable behaviors such as reflection and backtracking. Visualization and clustering analyses show that these behaviors occupy separable regions in the decoder column space. Moreover, targeted interventions on SAE-derived vectors can controllably amplify or suppress specific reasoning behaviors, altering inference trajectories without retraining. Beyond behavior-specific disentanglement, SAEs capture structural properties such as response length, revealing clusters of long versus short reasoning traces. More interestingly, SAEs enable the discovery of novel behaviors beyond human supervision. We demonstrate the ability to control response confidence by identifying confidence-related vectors in the SAE decoder space. These findings underscore the potential of unsupervised latent discovery for both interpreting and controllably steering reasoning in LLMs.</p> - https://arxiv.org/abs/2512.23988 - Tue, 30 Dec 2025 05:09:11 +0000 + One Sample to Rule Them All: Extreme Data Efficiency in RL Scaling + https://arxiv.org/abs/2601.03111 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03111.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yiyuan Li, Zhen Huang, Yanan Wu, Weixun Wang, Xuefeng Li, Yijia Luo, Wenbo Su, Bo Zheng, Pengfei Liu</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> The reasoning ability of large language models (LLMs) can be unleashed with reinforcement learning (RL) (OpenAI, 2024; DeepSeek-AI et al., 2025a; Zeng et al., 2025). The success of existing RL attempts in LLMs usually relies on high-quality samples of thousands or beyond. In this paper, we challenge fundamental assumptions about data requirements in RL for LLMs by demonstrating the remarkable effectiveness of one-shot learning. Specifically, we introduce polymath learning, a framework for designing one training sample that elicits multidisciplinary impact. We present three key findings: (1) A single, strategically selected math reasoning sample can produce significant performance improvements across multiple domains, including physics, chemistry, and biology with RL; (2) The math skills salient to reasoning suggest the characteristics of the optimal polymath sample; and (3) An engineered synthetic sample that integrates multidiscipline elements outperforms training with individual samples that naturally occur. Our approach achieves superior performance to training with larger datasets across various reasoning benchmarks, demonstrating that sample quality and design, rather than quantity, may be the key to unlock enhanced reasoning capabilities in language models. Our results suggest a shift, dubbed as sample engineering, toward precision engineering of training samples rather than simply increasing data volume.</p> + https://arxiv.org/abs/2601.03111 + Tue, 06 Jan 2026 15:41:35 +0000 - TESO Tabu Enhanced Simulation Optimization for Noisy Black Box Problems - https://arxiv.org/abs/2512.24007 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24007.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bulent Soykan, Sean Mondesire, Ghaith Rabadi</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Simulation optimization (SO) is frequently challenged by noisy evaluations, high computational costs, and complex, multimodal search landscapes. This paper introduces Tabu-Enhanced Simulation Optimization (TESO), a novel metaheuristic framework integrating adaptive search with memory-based strategies. TESO leverages a short-term Tabu List to prevent cycling and encourage diversification, and a long-term Elite Memory to guide intensification by perturbing high-performing solutions. An aspiration criterion allows overriding tabu restrictions for exceptional candidates. This combination facilitates a dynamic balance between exploration and exploitation in stochastic environments. We demonstrate TESO's effectiveness and reliability using an queue optimization problem, showing improved performance compared to benchmarks and validating the contribution of its memory components. Source code and data are available at: https://github.com/bulentsoykan/TESO.</p> - https://arxiv.org/abs/2512.24007 - Tue, 30 Dec 2025 06:03:37 +0000 + Unified Thinker: A General Reasoning Modular Core for Image Generation + https://arxiv.org/abs/2601.03127 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03127.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sashuai Zhou, Qiang Zhou, Jijin Hu, Hanqing Yang, Yue Cao, Junpeng Ma, Yinchao Ma, Jun Song, Tiezheng Ge, Cheng Yu, Bo Zheng, Zhou Zhao</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Despite impressive progress in high-fidelity image synthesis, generative models still struggle with logic-intensive instruction following, exposing a persistent reasoning--execution gap. Meanwhile, closed-source systems (e.g., Nano Banana) have demonstrated strong reasoning-driven image generation, highlighting a substantial gap to current open-source models. We argue that closing this gap requires not merely better visual generators, but executable reasoning: decomposing high-level intents into grounded, verifiable plans that directly steer the generative process. To this end, we propose Unified Thinker, a task-agnostic reasoning architecture for general image generation, designed as a unified planning core that can plug into diverse generators and workflows. Unified Thinker decouples a dedicated Thinker from the image Generator, enabling modular upgrades of reasoning without retraining the entire generative model. We further introduce a two-stage training paradigm: we first build a structured planning interface for the Thinker, then apply reinforcement learning to ground its policy in pixel-level feedback, encouraging plans that optimize visual correctness over textual plausibility. Extensive experiments on text-to-image generation and image editing show that Unified Thinker substantially improves image reasoning and generation quality.</p> + https://arxiv.org/abs/2601.03127 + Tue, 06 Jan 2026 15:59:33 +0000 - Factorized Learning for Temporally Grounded Video-Language Models - https://arxiv.org/abs/2512.24097 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24097.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wenzheng Zeng, Difei Gao, Mike Zheng Shou, Hwee Tou Ng</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Recent video-language models have shown great potential for video understanding, but still struggle with accurate temporal grounding for event-level perception. We observe that two main factors in video understanding (i.e., temporal grounding and textual response) form a logical hierarchy: accurate temporal evidence grounding lays the foundation for reliable textual response. However, existing works typically handle these two tasks in a coupled manner without a clear logical structure, leading to sub-optimal objectives. We address this from a factorized learning perspective. We first propose D^2VLM, a framework that decouples the learning of these two tasks while also emphasizing their inherent dependency. We adopt a "grounding then answering with evidence referencing" paradigm and introduce evidence tokens for evidence grounding, which emphasize event-level visual semantic capture beyond the focus on timestamp representation in existing works. To further facilitate the learning of these two tasks, we introduce a novel factorized preference optimization (FPO) algorithm. Unlike standard preference optimization, FPO explicitly incorporates probabilistic temporal grounding modeling into the optimization objective, enabling preference learning for both temporal grounding and textual response. We also construct a synthetic dataset to address the lack of suitable datasets for factorized preference learning with explicit temporal grounding. Experiments on various tasks demonstrate the clear advantage of our approach. Our source code is available at https://github.com/nusnlp/d2vlm.</p> - https://arxiv.org/abs/2512.24097 - Tue, 30 Dec 2025 09:13:20 +0000 + Parallel Latent Reasoning for Sequential Recommendation + https://arxiv.org/abs/2601.03153 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03153.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiakai Tang, Xu Chen, Wen Chen, Jian Wu, Yuning Jiang, Bo Zheng</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Capturing complex user preferences from sparse behavioral sequences remains a fundamental challenge in sequential recommendation. Recent latent reasoning methods have shown promise by extending test-time computation through multi-step reasoning, yet they exclusively rely on depth-level scaling along a single trajectory, suffering from diminishing returns as reasoning depth increases. To address this limitation, we propose Parallel Latent Reasoning (PLR), a novel framework that pioneers width-level computational scaling by exploring multiple diverse reasoning trajectories simultaneously. PLR constructs parallel reasoning streams through learnable trigger tokens in continuous latent space, preserves diversity across streams via global reasoning regularization, and adaptively synthesizes multi-stream outputs through mixture-of-reasoning-streams aggregation. Extensive experiments on three real-world datasets demonstrate that PLR substantially outperforms state-of-the-art baselines while maintaining real-time inference efficiency. Theoretical analysis further validates the effectiveness of parallel reasoning in improving generalization capability. Our work opens new avenues for enhancing reasoning capacity in sequential recommendation beyond existing depth scaling.</p> + https://arxiv.org/abs/2601.03153 + Tue, 06 Jan 2026 16:25:48 +0000 - DiffThinker: Towards Generative Multimodal Reasoning with Diffusion Models - https://arxiv.org/abs/2512.24165 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24165.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zefeng He, Xiaoye Qu, Yafu Li, Tong Zhu, Siyuan Huang, Yu Cheng</p><p><b>Upvotes:</b> 41</p><p><b>Summary:</b> While recent Multimodal Large Language Models (MLLMs) have attained significant strides in multimodal reasoning, their reasoning processes remain predominantly text-centric, leading to suboptimal performance in complex long-horizon, vision-centric tasks. In this paper, we establish a novel Generative Multimodal Reasoning paradigm and introduce DiffThinker, a diffusion-based reasoning framework. Conceptually, DiffThinker reformulates multimodal reasoning as a native generative image-to-image task, achieving superior logical consistency and spatial precision in vision-centric tasks. We perform a systematic comparison between DiffThinker and MLLMs, providing the first in-depth investigation into the intrinsic characteristics of this paradigm, revealing four core properties: efficiency, controllability, native parallelism, and collaboration. Extensive experiments across four domains (sequential planning, combinatorial optimization, constraint satisfaction, and spatial configuration) demonstrate that DiffThinker significantly outperforms leading closed source models including GPT-5 (+314.2\%) and Gemini-3-Flash (+111.6\%), as well as the fine-tuned Qwen3-VL-32B baseline (+39.0\%), highlighting generative multimodal reasoning as a promising approach for vision-centric reasoning.</p> - https://arxiv.org/abs/2512.24165 - Tue, 30 Dec 2025 11:51:18 +0000 + UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision + https://arxiv.org/abs/2601.03193 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03193.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ruiyan Han, Zhen Fang, XinYu Sun, Yuchen Ma, Ziheng Wang, Yu Zeng, Zehui Chen, Lin Chen, Wenxuan Huang, Wei-Jie Xu, Yi Cao, Feng Zhao</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> While Unified Multimodal Models (UMMs) have achieved remarkable success in cross-modal comprehension, a significant gap persists in their ability to leverage such internal knowledge for high-quality generation. We formalize this discrepancy as Conduction Aphasia, a phenomenon where models accurately interpret multimodal inputs but struggle to translate that understanding into faithful and controllable synthesis. To address this, we propose UniCorn, a simple yet elegant self-improvement framework that eliminates the need for external data or teacher supervision. By partitioning a single UMM into three collaborative roles: Proposer, Solver, and Judge, UniCorn generates high-quality interactions via self-play and employs cognitive pattern reconstruction to distill latent understanding into explicit generative signals. To validate the restoration of multimodal coherence, we introduce UniCycle, a cycle-consistency benchmark based on a Text to Image to Text reconstruction loop. Extensive experiments demonstrate that UniCorn achieves comprehensive and substantial improvements over the base model across six general image generation benchmarks. Notably, it achieves SOTA performance on TIIF(73.8), DPG(86.8), CompBench(88.5), and UniCycle while further delivering substantial gains of +5.0 on WISE and +6.5 on OneIG. These results highlight that our method significantly enhances T2I generation while maintaining robust comprehension, demonstrating the scalability of fully self-supervised refinement for unified multimodal intelligence.</p> + https://arxiv.org/abs/2601.03193 + Tue, 06 Jan 2026 17:15:50 +0000 - Guiding a Diffusion Transformer with the Internal Dynamics of Itself - https://arxiv.org/abs/2512.24176 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24176.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xingyu Zhou, Qifan Li, Xiaobin Hu, Hai Chen, Shuhang Gu</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> The diffusion model presents a powerful ability to capture the entire (conditional) data distribution. However, due to the lack of sufficient training and data to learn to cover low-probability areas, the model will be penalized for failing to generate high-quality images corresponding to these areas. To achieve better generation quality, guidance strategies such as classifier free guidance (CFG) can guide the samples to the high-probability areas during the sampling stage. However, the standard CFG often leads to over-simplified or distorted samples. On the other hand, the alternative line of guiding diffusion model with its bad version is limited by carefully designed degradation strategies, extra training and additional sampling steps. In this paper, we proposed a simple yet effective strategy Internal Guidance (IG), which introduces an auxiliary supervision on the intermediate layer during training process and extrapolates the intermediate and deep layer's outputs to obtain generative results during sampling process. This simple strategy yields significant improvements in both training efficiency and generation quality on various baselines. On ImageNet 256x256, SiT-XL/2+IG achieves FID=5.31 and FID=1.75 at 80 and 800 epochs. More impressively, LightningDiT-XL/1+IG achieves FID=1.34 which achieves a large margin between all of these methods. Combined with CFG, LightningDiT-XL/1+IG achieves the current state-of-the-art FID of 1.19.</p> - https://arxiv.org/abs/2512.24176 - Tue, 30 Dec 2025 12:16:46 +0000 + LTX-2: Efficient Joint Audio-Visual Foundation Model + https://arxiv.org/abs/2601.03233 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03233.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yoav HaCohen, Benny Brazowski, Nisan Chiprut, Yaki Bitterman, Andrew Kvochko, Avishai Berkowitz, Daniel Shalem, Daphna Lifschitz, Dudu Moshe, Eitan Porat, Eitan Richardson, Guy Shiran, Itay Chachy, Jonathan Chetboun, Michael Finkelson, Michael Kupchick, Nir Zabari, Nitzan Guetta, Noa Kotler, Ofir Bibi, Ori Gordon, Poriya Panet, Roi Benita, Shahar Armon, Victor Kulikov, Yaron Inger, Yonatan Shiftan, Zeev Melumian, Zeev Farbman</p><p><b>Upvotes:</b> 88</p><p><b>Summary:</b> Recent text-to-video diffusion models can generate compelling video sequences, yet they remain silent -- missing the semantic, emotional, and atmospheric cues that audio provides. We introduce LTX-2, an open-source foundational model capable of generating high-quality, temporally synchronized audiovisual content in a unified manner. LTX-2 consists of an asymmetric dual-stream transformer with a 14B-parameter video stream and a 5B-parameter audio stream, coupled through bidirectional audio-video cross-attention layers with temporal positional embeddings and cross-modality AdaLN for shared timestep conditioning. This architecture enables efficient training and inference of a unified audiovisual model while allocating more capacity for video generation than audio generation. We employ a multilingual text encoder for broader prompt understanding and introduce a modality-aware classifier-free guidance (modality-CFG) mechanism for improved audiovisual alignment and controllability. Beyond generating speech, LTX-2 produces rich, coherent audio tracks that follow the characters, environment, style, and emotion of each scene -- complete with natural background and foley elements. In our evaluations, the model achieves state-of-the-art audiovisual quality and prompt adherence among open-source systems, while delivering results comparable to proprietary models at a fraction of their computational cost and inference time. All model weights and code are publicly released.</p> + https://arxiv.org/abs/2601.03233 + Tue, 06 Jan 2026 18:24:41 +0000 - GR-Dexter Technical Report - https://arxiv.org/abs/2512.24210 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24210.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ruoshi Wen, Guangzeng Chen, Zhongren Cui, Min Du, Yang Gou, Zhigang Han, Liqun Huang, Mingyu Lei, Yunfei Li, Zhuohang Li, Wenlei Liu, Yuxiao Liu, Xiao Ma, Hao Niu, Yutao Ouyang, Zeyu Ren, Haixin Shi, Wei Xu, Haoxiang Zhang, Jiajun Zhang, Xiao Zhang, Liwei Zheng, Weiheng Zhong, Yifei Zhou, Zhengming Zhu, Hang Li</p><p><b>Upvotes:</b> 19</p><p><b>Summary:</b> Vision-language-action (VLA) models have enabled language-conditioned, long-horizon robot manipulation, but most existing systems are limited to grippers. Scaling VLA policies to bimanual robots with high degree-of-freedom (DoF) dexterous hands remains challenging due to the expanded action space, frequent hand-object occlusions, and the cost of collecting real-robot data. We present GR-Dexter, a holistic hardware-model-data framework for VLA-based generalist manipulation on a bimanual dexterous-hand robot. Our approach combines the design of a compact 21-DoF robotic hand, an intuitive bimanual teleoperation system for real-robot data collection, and a training recipe that leverages teleoperated robot trajectories together with large-scale vision-language and carefully curated cross-embodiment datasets. Across real-world evaluations spanning long-horizon everyday manipulation and generalizable pick-and-place, GR-Dexter achieves strong in-domain performance and improved robustness to unseen objects and unseen instructions. We hope GR-Dexter serves as a practical step toward generalist dexterous-hand robotic manipulation.</p> - https://arxiv.org/abs/2512.24210 - Tue, 30 Dec 2025 13:22:16 +0000 + MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents + https://arxiv.org/abs/2601.03236 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03236.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dongming Jiang, Yi Li, Guanpeng Li, Bingzhe Li</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Memory-Augmented Generation (MAG) extends Large Language Models with external memory to support long-context reasoning, but existing approaches largely rely on semantic similarity over monolithic memory stores, entangling temporal, causal, and entity information. This design limits interpretability and alignment between query intent and retrieved evidence, leading to suboptimal reasoning accuracy. In this paper, we propose MAGMA, a multi-graph agentic memory architecture that represents each memory item across orthogonal semantic, temporal, causal, and entity graphs. MAGMA formulates retrieval as policy-guided traversal over these relational views, enabling query-adaptive selection and structured context construction. By decoupling memory representation from retrieval logic, MAGMA provides transparent reasoning paths and fine-grained control over retrieval. Experiments on LoCoMo and LongMemEval demonstrate that MAGMA consistently outperforms state-of-the-art agentic memory systems in long-horizon reasoning tasks.</p> + https://arxiv.org/abs/2601.03236 + Tue, 06 Jan 2026 18:29:43 +0000 - Figure It Out: Improving the Frontier of Reasoning with Active Visual Thinking - https://arxiv.org/abs/2512.24297 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24297.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Meiqi Chen, Fandong Meng, Jie Zhou</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Complex reasoning problems often involve implicit spatial, geometric, and structural relationships that are not explicitly encoded in text. While recent reasoning models have achieved strong performance across many domains, purely text-based reasoning struggles to represent global structural constraints in complex settings. In this paper, we introduce FIGR, which integrates active visual thinking into multi-turn reasoning via end-to-end reinforcement learning. FIGR externalizes intermediate structural hypotheses by constructing visual representations during problem solving. By adaptively regulating when and how visual reasoning should be invoked, FIGR enables more stable and coherent reasoning over global structural properties that are difficult to capture from text alone. Experiments on challenging mathematical reasoning benchmarks demonstrate that FIGR outperforms strong text-only chain-of-thought baselines. In particular, FIGR improves the base model by 13.12% on AIME 2025 and 11.00% on BeyondAIME, highlighting the effectiveness of figure-guided multimodal reasoning in enhancing the stability and reliability of complex reasoning.</p> - https://arxiv.org/abs/2512.24297 - Tue, 30 Dec 2025 15:39:11 +0000 + InfiniDepth: Arbitrary-Resolution and Fine-Grained Depth Estimation with Neural Implicit Fields + https://arxiv.org/abs/2601.03252 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03252.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hao Yu, Haotong Lin, Jiawei Wang, Jiaxin Li, Yida Wang, Xueyang Zhang, Yue Wang, Xiaowei Zhou, Ruizhen Hu, Sida Peng</p><p><b>Upvotes:</b> 94</p><p><b>Summary:</b> Existing depth estimation methods are fundamentally limited to predicting depth on discrete image grids. Such representations restrict their scalability to arbitrary output resolutions and hinder the geometric detail recovery. This paper introduces InfiniDepth, which represents depth as neural implicit fields. Through a simple yet effective local implicit decoder, we can query depth at continuous 2D coordinates, enabling arbitrary-resolution and fine-grained depth estimation. To better assess our method's capabilities, we curate a high-quality 4K synthetic benchmark from five different games, spanning diverse scenes with rich geometric and appearance details. Extensive experiments demonstrate that InfiniDepth achieves state-of-the-art performance on both synthetic and real-world benchmarks across relative and metric depth estimation tasks, particularly excelling in fine-detail regions. It also benefits the task of novel view synthesis under large viewpoint shifts, producing high-quality results with fewer holes and artifacts.</p> + https://arxiv.org/abs/2601.03252 + Tue, 06 Jan 2026 18:57:06 +0000 - Forging Spatial Intelligence: A Roadmap of Multi-Modal Data Pre-Training for Autonomous Systems - https://arxiv.org/abs/2512.24385 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24385.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Song Wang, Lingdong Kong, Xiaolu Liu, Hao Shi, Wentong Li, Jianke Zhu, Steven C. H. Hoi</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> The rapid advancement of autonomous systems, including self-driving vehicles and drones, has intensified the need to forge true Spatial Intelligence from multi-modal onboard sensor data. While foundation models excel in single-modal contexts, integrating their capabilities across diverse sensors like cameras and LiDAR to create a unified understanding remains a formidable challenge. This paper presents a comprehensive framework for multi-modal pre-training, identifying the core set of techniques driving progress toward this goal. We dissect the interplay between foundational sensor characteristics and learning strategies, evaluating the role of platform-specific datasets in enabling these advancements. Our central contribution is the formulation of a unified taxonomy for pre-training paradigms: ranging from single-modality baselines to sophisticated unified frameworks that learn holistic representations for advanced tasks like 3D object detection and semantic occupancy prediction. Furthermore, we investigate the integration of textual inputs and occupancy representations to facilitate open-world perception and planning. Finally, we identify critical bottlenecks, such as computational efficiency and model scalability, and propose a roadmap toward general-purpose multi-modal foundation models capable of achieving robust Spatial Intelligence for real-world deployment.</p> - https://arxiv.org/abs/2512.24385 - Tue, 30 Dec 2025 17:58:01 +0000 + Muses: Designing, Composing, Generating Nonexistent Fantasy 3D Creatures without Training + https://arxiv.org/abs/2601.03256 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03256.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hexiao Lu, Xiaokun Sun, Zeyu Cai, Hao Guo, Ying Tai, Jian Yang, Zhenyu Zhang</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> We present Muses, the first training-free method for fantastic 3D creature generation in a feed-forward paradigm. Previous methods, which rely on part-aware optimization, manual assembly, or 2D image generation, often produce unrealistic or incoherent 3D assets due to the challenges of intricate part-level manipulation and limited out-of-domain generation. In contrast, Muses leverages the 3D skeleton, a fundamental representation of biological forms, to explicitly and rationally compose diverse elements. This skeletal foundation formalizes 3D content creation as a structure-aware pipeline of design, composition, and generation. Muses begins by constructing a creatively composed 3D skeleton with coherent layout and scale through graph-constrained reasoning. This skeleton then guides a voxel-based assembly process within a structured latent space, integrating regions from different objects. Finally, image-guided appearance modeling under skeletal conditions is applied to generate a style-consistent and harmonious texture for the assembled shape. Extensive experiments establish Muses' state-of-the-art performance in terms of visual fidelity and alignment with textual descriptions, and potential on flexible 3D object editing. Project page: https://luhexiao.github.io/Muses.github.io/.</p> + https://arxiv.org/abs/2601.03256 + Tue, 06 Jan 2026 18:59:57 +0000 - PhyGDPO: Physics-Aware Groupwise Direct Preference Optimization for Physically Consistent Text-to-Video Generation - https://arxiv.org/abs/2512.24551 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24551.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuanhao Cai, Kunpeng Li, Menglin Jia, Jialiang Wang, Junzhe Sun, Feng Liang, Weifeng Chen, Felix Juefei-Xu, Chu Wang, Ali Thabet, Xiaoliang Dai, Xuan Ju, Alan Yuille, Ji Hou</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Recent advances in text-to-video (T2V) generation have achieved good visual quality, yet synthesizing videos that faithfully follow physical laws remains an open challenge. Existing methods mainly based on graphics or prompt extension struggle to generalize beyond simple simulated environments or learn implicit physical reasoning. The scarcity of training data with rich physics interactions and phenomena is also a problem. In this paper, we first introduce a Physics-Augmented video data construction Pipeline, PhyAugPipe, that leverages a vision-language model (VLM) with chain-of-thought reasoning to collect a large-scale training dataset, PhyVidGen-135K. Then we formulate a principled Physics-aware Groupwise Direct Preference Optimization, PhyGDPO, framework that builds upon the groupwise Plackett-Luce probabilistic model to capture holistic preferences beyond pairwise comparisons. In PhyGDPO, we design a Physics-Guided Rewarding (PGR) scheme that embeds VLM-based physics rewards to steer optimization toward physical consistency. We also propose a LoRA-Switch Reference (LoRA-SR) scheme that eliminates memory-heavy reference duplication for efficient training. Experiments show that our method significantly outperforms state-of-the-art open-source methods on PhyGenBench and VideoPhy2. Please check our project page at https://caiyuanhao1998.github.io/project/PhyGDPO for more video results. Our code, models, and data will be released at https://github.com/caiyuanhao1998/Open-PhyGDPO</p> - https://arxiv.org/abs/2512.24551 - Wed, 31 Dec 2025 01:19:14 +0000 + Guardians of the Hair: Rescuing Soft Boundaries in Depth, Stereo, and Novel Views + https://arxiv.org/abs/2601.03362 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03362.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiang Zhang, Yang Zhang, Lukas Mehl, Markus Gross, Christopher Schroers</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Soft boundaries, like thin hairs, are commonly observed in natural and computer-generated imagery, but they remain challenging for 3D vision due to the ambiguous mixing of foreground and background cues. This paper introduces Guardians of the Hair (HairGuard), a framework designed to recover fine-grained soft boundary details in 3D vision tasks. Specifically, we first propose a novel data curation pipeline that leverages image matting datasets for training and design a depth fixer network to automatically identify soft boundary regions. With a gated residual module, the depth fixer refines depth precisely around soft boundaries while maintaining global depth quality, allowing plug-and-play integration with state-of-the-art depth models. For view synthesis, we perform depth-based forward warping to retain high-fidelity textures, followed by a generative scene painter that fills disoccluded regions and eliminates redundant background artifacts within soft boundaries. Finally, a color fuser adaptively combines warped and inpainted results to produce novel views with consistent geometry and fine-grained details. Extensive experiments demonstrate that HairGuard achieves state-of-the-art performance across monocular depth estimation, stereo image/video conversion, and novel view synthesis, with significant improvements in soft boundary regions.</p> + https://arxiv.org/abs/2601.03362 + Tue, 06 Jan 2026 19:02:34 +0000 - Dynamic Large Concept Models: Latent Reasoning in an Adaptive Semantic Space - https://arxiv.org/abs/2512.24617 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24617.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xingwei Qu, Shaowen Wang, Zihao Huang, Kai Hua, Fan Yin, Rui-Jie Zhu, Jundong Zhou, Qiyang Min, Zihao Wang, Yizhi Li, Tianyu Zhang, He Xing, Zheng Zhang, Yuxuan Song, Tianyu Zheng, Zhiyuan Zeng, Chenghua Lin, Ge Zhang, Wenhao Huang</p><p><b>Upvotes:</b> 44</p><p><b>Summary:</b> Large Language Models (LLMs) apply uniform computation to all tokens, despite language exhibiting highly non-uniform information density. This token-uniform regime wastes capacity on locally predictable spans while under-allocating computation to semantically critical transitions. We propose Dynamic Large Concept Models (DLCM), a hierarchical language modeling framework that learns semantic boundaries from latent representations and shifts computation from tokens to a compressed concept space where reasoning is more efficient. DLCM discovers variable-length concepts end-to-end without relying on predefined linguistic units. Hierarchical compression fundamentally changes scaling behavior. We introduce the first compression-aware scaling law, which disentangles token-level capacity, concept-level reasoning capacity, and compression ratio, enabling principled compute allocation under fixed FLOPs. To stably train this heterogeneous architecture, we further develop a decoupled μP parametrization that supports zero-shot hyperparameter transfer across widths and compression regimes. At a practical setting (R=4, corresponding to an average of four tokens per concept), DLCM reallocates roughly one-third of inference compute into a higher-capacity reasoning backbone, achieving a +2.69\% average improvement across 12 zero-shot benchmarks under matched inference FLOPs.</p> - https://arxiv.org/abs/2512.24617 - Wed, 31 Dec 2025 04:19:33 +0000 + The Illusion of Specialization: Unveiling the Domain-Invariant "Standing Committee" in Mixture-of-Experts Models + https://arxiv.org/abs/2601.03425 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03425.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yan Wang, Yitao Xu, Nanhan Shen, Jinyan Su, Jimin Huang, Zining Zhu</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> Mixture of Experts models are widely assumed to achieve domain specialization through sparse routing. In this work, we question this assumption by introducing COMMITTEEAUDIT, a post hoc framework that analyzes routing behavior at the level of expert groups rather than individual experts. Across three representative models and the MMLU benchmark, we uncover a domain-invariant Standing Committee. This is a compact coalition of routed experts that consistently captures the majority of routing mass across domains, layers, and routing budgets, even when architectures already include shared experts. Qualitative analysis further shows that Standing Committees anchor reasoning structure and syntax, while peripheral experts handle domain-specific knowledge. These findings reveal a strong structural bias toward centralized computation, suggesting that specialization in Mixture of Experts models is far less pervasive than commonly believed. This inherent bias also indicates that current training objectives, such as load-balancing losses that enforce uniform expert utilization, may be working against the model's natural optimization path, thereby limiting training efficiency and performance.</p> + https://arxiv.org/abs/2601.03425 + Tue, 06 Jan 2026 21:29:45 +0000 - Youtu-LLM: Unlocking the Native Agentic Potential for Lightweight Large Language Models - https://arxiv.org/abs/2512.24618 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24618.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Junru Lu, Jiarui Qin, Lingfeng Qiao, Yinghui Li, Xinyi Dai, Bo Ke, Jianfeng He, Ruizhi Qiao, Di Yin, Xing Sun, Yunsheng Wu, Yinsong Liu, Shuangyin Liu, Mingkong Tang, Haodong Lin, Jiayi Kuang, Fanxu Meng, Xiaojuan Tang, Yunjia Xi, Junjie Huang, Haotong Yang, Zhenyi Shen, Yangning Li, Qianwen Zhang, Yifei Yu, Siyu An, Junnan Dong, Qiufeng Wang, Jie Wang, Keyu Chen, Wei Wen, Taian Guo, Zhifeng Shen, Daohai Yu, Jiahao Li, Ke Li, Zongyi Li, Xiaoyu Tan</p><p><b>Upvotes:</b> 104</p><p><b>Summary:</b> We introduce Youtu-LLM, a lightweight yet powerful language model that harmonizes high computational efficiency with native agentic intelligence. Unlike typical small models that rely on distillation, Youtu-LLM (1.96B) is pre-trained from scratch to systematically cultivate reasoning and planning capabilities. The key technical advancements are as follows: (1) Compact Architecture with Long-Context Support: Built on a dense Multi-Latent Attention (MLA) architecture with a novel STEM-oriented vocabulary, Youtu-LLM supports a 128k context window. This design enables robust long-context reasoning and state tracking within a minimal memory footprint, making it ideal for long-horizon agent and reasoning tasks. (2) Principled "Commonsense-STEM-Agent" Curriculum: We curated a massive corpus of approximately 11T tokens and implemented a multi-stage training strategy. By progressively shifting the pre-training data distribution from general commonsense to complex STEM and agentic tasks, we ensure the model acquires deep cognitive abilities rather than superficial alignment. (3) Scalable Agentic Mid-training: Specifically for the agentic mid-training, we employ diverse data construction schemes to synthesize rich and varied trajectories across math, coding, and tool-use domains. This high-quality data enables the model to internalize planning and reflection behaviors effectively. Extensive evaluations show that Youtu-LLM sets a new state-of-the-art for sub-2B LLMs. On general benchmarks, it achieves competitive performance against larger models, while on agent-specific tasks, it significantly surpasses existing SOTA baselines, demonstrating that lightweight models can possess strong intrinsic agentic capabilities.</p> - https://arxiv.org/abs/2512.24618 - Wed, 31 Dec 2025 04:25:11 +0000 + Enhancing Linguistic Competence of Language Models through Pre-training with Language Learning Tasks + https://arxiv.org/abs/2601.03448 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03448.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Atsuki Yamaguchi, Maggie Mi, Nikolaos Aletras</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Language models (LMs) are pre-trained on raw text datasets to generate text sequences token-by-token. While this approach facilitates the learning of world knowledge and reasoning, it does not explicitly optimize for linguistic competence. To bridge this gap, we propose L2T, a pre-training framework integrating Language Learning Tasks alongside standard next-token prediction. Inspired by human language acquisition, L2T transforms raw text into structured input-output pairs to provide explicit linguistic stimulation. Pre-training LMs on a mixture of raw text and L2T data not only improves overall performance on linguistic competence benchmarks but accelerates its acquisition, while maintaining competitive performance on general reasoning tasks.</p> + https://arxiv.org/abs/2601.03448 + Tue, 06 Jan 2026 22:28:15 +0000 - FlowBlending: Stage-Aware Multi-Model Sampling for Fast and High-Fidelity Video Generation - https://arxiv.org/abs/2512.24724 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24724.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jibin Song, Mingi Kwon, Jaeseok Jeong, Youngjung Uh</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> In this work, we show that the impact of model capacity varies across timesteps: it is crucial for the early and late stages but largely negligible during the intermediate stage. Accordingly, we propose FlowBlending, a stage-aware multi-model sampling strategy that employs a large model and a small model at capacity-sensitive stages and intermediate stages, respectively. We further introduce simple criteria to choose stage boundaries and provide a velocity-divergence analysis as an effective proxy for identifying capacity-sensitive regions. Across LTX-Video (2B/13B) and WAN 2.1 (1.3B/14B), FlowBlending achieves up to 1.65x faster inference with 57.35% fewer FLOPs, while maintaining the visual fidelity, temporal coherence, and semantic alignment of the large models. FlowBlending is also compatible with existing sampling-acceleration techniques, enabling up to 2x additional speedup. Project page is available at: https://jibin86.github.io/flowblending_project_page.</p> - https://arxiv.org/abs/2512.24724 - Wed, 31 Dec 2025 08:41:27 +0000 + ThinkRL-Edit: Thinking in Reinforcement Learning for Reasoning-Centric Image Editing + https://arxiv.org/abs/2601.03467 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03467.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hengjia Li, Liming Jiang, Qing Yan, Yizhi Song, Hao Kang, Zichuan Liu, Xin Lu, Boxi Wu, Deng Cai</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Instruction-driven image editing with unified multimodal generative models has advanced rapidly, yet their underlying visual reasoning remains limited, leading to suboptimal performance on reasoning-centric edits. Reinforcement learning (RL) has been investigated for improving the quality of image editing, but it faces three key challenges: (1) limited reasoning exploration confined to denoising stochasticity, (2) biased reward fusion, and (3) unstable VLM-based instruction rewards. In this work, we propose ThinkRL-Edit, a reasoning-centric RL framework that decouples visual reasoning from image synthesis and expands reasoning exploration beyond denoising. To the end, we introduce Chain-of-Thought (CoT)-based reasoning sampling with planning and reflection stages prior to generation in online sampling, compelling the model to explore multiple semantic hypotheses and validate their plausibility before committing to a visual outcome. To avoid the failures of weighted aggregation, we propose an unbiased chain preference grouping strategy across multiple reward dimensions. Moreover, we replace interval-based VLM scores with a binary checklist, yielding more precise, lower-variance, and interpretable rewards for complex reasoning. Experiments show our method significantly outperforms prior work on reasoning-centric image editing, producing instruction-faithful, visually coherent, and semantically grounded edits.</p> + https://arxiv.org/abs/2601.03467 + Tue, 06 Jan 2026 23:43:00 +0000 - Dream2Flow: Bridging Video Generation and Open-World Manipulation with 3D Object Flow - https://arxiv.org/abs/2512.24766 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24766.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Karthik Dharmarajan, Wenlong Huang, Jiajun Wu, Li Fei-Fei, Ruohan Zhang</p><p><b>Upvotes:</b> 6</p><p><b>Summary:</b> Generative video modeling has emerged as a compelling tool to zero-shot reason about plausible physical interactions for open-world manipulation. Yet, it remains a challenge to translate such human-led motions into the low-level actions demanded by robotic systems. We observe that given an initial image and task instruction, these models excel at synthesizing sensible object motions. Thus, we introduce Dream2Flow, a framework that bridges video generation and robotic control through 3D object flow as an intermediate representation. Our method reconstructs 3D object motions from generated videos and formulates manipulation as object trajectory tracking. By separating the state changes from the actuators that realize those changes, Dream2Flow overcomes the embodiment gap and enables zero-shot guidance from pre-trained video models to manipulate objects of diverse categories-including rigid, articulated, deformable, and granular. Through trajectory optimization or reinforcement learning, Dream2Flow converts reconstructed 3D object flow into executable low-level commands without task-specific demonstrations. Simulation and real-world experiments highlight 3D object flow as a general and scalable interface for adapting video generation models to open-world robotic manipulation. Videos and visualizations are available at https://dream2flow.github.io/.</p> - https://arxiv.org/abs/2512.24766 - Wed, 31 Dec 2025 10:25:24 +0000 + EpiQAL: Benchmarking Large Language Models in Epidemiological Question Answering for Enhanced Alignment and Reasoning + https://arxiv.org/abs/2601.03471 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03471.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingyang Wei, Dehai Min, Zewen Liu, Yuzhang Xie, Guanchen Wu, Carl Yang, Max S. Y. Lau, Qi He, Lu Cheng, Wei Jin</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Reliable epidemiological reasoning requires synthesizing study evidence to infer disease burden, transmission dynamics, and intervention effects at the population level. Existing medical question answering benchmarks primarily emphasize clinical knowledge or patient-level reasoning, yet few systematically evaluate evidence-grounded epidemiological inference. We present EpiQAL, the first diagnostic benchmark for epidemiological question answering across diverse diseases, comprising three subsets built from open-access literature. The subsets respectively evaluate text-grounded factual recall, multi-step inference linking document evidence with epidemiological principles, and conclusion reconstruction with the Discussion section withheld. Construction combines expert-designed taxonomy guidance, multi-model verification, and retrieval-based difficulty control. Experiments on ten open models reveal that current LLMs show limited performance on epidemiological reasoning, with multi-step inference posing the greatest challenge. Model rankings shift across subsets, and scale alone does not predict success. Chain-of-Thought prompting benefits multi-step inference but yields mixed results elsewhere. EpiQAL provides fine-grained diagnostic signals for evidence grounding, inferential reasoning, and conclusion reconstruction.</p> + https://arxiv.org/abs/2601.03471 + Tue, 06 Jan 2026 23:49:10 +0000 - Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem - https://arxiv.org/abs/2512.24873 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24873.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Weixun Wang, XiaoXiao Xu, Wanhe An, Fangwen Dai, Wei Gao, Yancheng He, Ju Huang, Qiang Ji, Hanqi Jin, Xiaoyang Li, Yang Li, Zhongwen Li, Shirong Lin, Jiashun Liu, Zenan Liu, Tao Luo, Dilxat Muhtar, Yuanbin Qu, Jiaqiang Shi, Qinghui Sun, Yingshui Tan, Hao Tang, Runze Wang, Yi Wang, Zhaoguo Wang, Yanan Wu, Shaopan Xiong, Binchen Xu, Xander Xu, Yuchi Xu, Qipeng Zhang, Xixia Zhang, Haizhou Zhao, Jie Zhao, Shuaibing Zhao, Baihui Zheng, Jianhui Zheng, Suhang Zheng, Yanni Zhu, Mengze Cai, Kerui Cao, Xitong Chen, Yue Dai, Lifan Du, Tao Feng, Tao He, Jin Hu, Yijie Hu, Ziyu Jiang, Cheng Li, Xiang Li, Jing Liang, Chonghuan Liu, ZhenDong Liu, Haodong Mi, Yanhu Mo, Junjia Ni, Shixin Pei, Jingyu Shen, XiaoShuai Song, Cecilia Wang, Chaofan Wang, Kangyu Wang, Pei Wang, Tao Wang, Wei Wang, Ke Xiao, Mingyu Xu, Tiange Xu, Nan Ya, Siran Yang, Jianan Ye, Yaxing Zang, Duo Zhang, Junbo Zhang, Boren Zheng, Wanxi Deng, Ling Pan, Lin Qu, Wenbo Su, Jiamang Wang, Wei Wang, Hu Wei, Minggang Wu, Cheng Yu, Bing Zhao, Zhicheng Zheng, Bo Zheng</p><p><b>Upvotes:</b> 51</p><p><b>Summary:</b> Agentic crafting requires LLMs to operate in real-world environments over multiple turns by taking actions, observing outcomes, and iteratively refining artifacts. Despite its importance, the open-source community lacks a principled, end-to-end ecosystem to streamline agent development. We introduce the Agentic Learning Ecosystem (ALE), a foundational infrastructure that optimizes the production pipeline for agent LLMs. ALE consists of three components: ROLL, a post-training framework for weight optimization; ROCK, a sandbox environment manager for trajectory generation; and iFlow CLI, an agent framework for efficient context engineering. We release ROME (ROME is Obviously an Agentic Model), an open-source agent grounded by ALE and trained on over one million trajectories. Our approach includes data composition protocols for synthesizing complex behaviors and a novel policy optimization algorithm, Interaction-based Policy Alignment (IPA), which assigns credit over semantic interaction chunks rather than individual tokens to improve long-horizon training stability. Empirically, we evaluate ROME within a structured setting and introduce Terminal Bench Pro, a benchmark with improved scale and contamination control. ROME demonstrates strong performance across benchmarks like SWE-bench Verified and Terminal Bench, proving the effectiveness of the ALE infrastructure.</p> - https://arxiv.org/abs/2512.24873 - Wed, 31 Dec 2025 14:03:39 +0000 + Evolving Programmatic Skill Networks + https://arxiv.org/abs/2601.03509 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03509.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haochen Shi, Xingdi Yuan, Bang Liu</p><p><b>Upvotes:</b> 71</p><p><b>Summary:</b> We study continual skill acquisition in open-ended embodied environments where an agent must construct, refine, and reuse an expanding library of executable skills. We introduce the Programmatic Skill Network (PSN), a framework in which skills are executable symbolic programs forming a compositional network that evolves through experience. PSN defines three core mechanisms instantiated via large language models: (1)REFLECT for structured fault localization over skill compositions, (2) progressive optimization with maturity-aware update gating that stabilizes reliable skills while maintaining plasticity for uncertain ones, and (3) canonical structural refactoring under rollback validation that maintains network compactness. We further show that PSN's learning dynamics exhibit structural parallels to neural network training. Experiments on MineDojo and Crafter demonstrate robust skill reuse, rapid adaptation, and strong generalization across open-ended task distributions.\footnote{We plan to open-source the code.</p> + https://arxiv.org/abs/2601.03509 + Wed, 07 Jan 2026 01:43:25 +0000 - mHC: Manifold-Constrained Hyper-Connections - https://arxiv.org/abs/2512.24880 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24880.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhenda Xie, Yixuan Wei, Huanqi Cao, Chenggang Zhao, Chengqi Deng, Jiashi Li, Damai Dai, Huazuo Gao, Jiang Chang, Liang Zhao, Shangyan Zhou, Zhean Xu, Zhengyan Zhang, Wangding Zeng, Shengding Hu, Yuqing Wang, Jingyang Yuan, Lean Wang, Wenfeng Liang</p><p><b>Upvotes:</b> 178</p><p><b>Summary:</b> Recently, studies exemplified by Hyper-Connections (HC) have extended the ubiquitous residual connection paradigm established over the past decade by expanding the residual stream width and diversifying connectivity patterns. While yielding substantial performance gains, this diversification fundamentally compromises the identity mapping property intrinsic to the residual connection, which causes severe training instability and restricted scalability, and additionally incurs notable memory access overhead. To address these challenges, we propose Manifold-Constrained Hyper-Connections (mHC), a general framework that projects the residual connection space of HC onto a specific manifold to restore the identity mapping property, while incorporating rigorous infrastructure optimization to ensure efficiency. Empirical experiments demonstrate that mHC is effective for training at scale, offering tangible performance improvements and superior scalability. We anticipate that mHC, as a flexible and practical extension of HC, will contribute to a deeper understanding of topological architecture design and suggest promising directions for the evolution of foundational models.</p> - https://arxiv.org/abs/2512.24880 - Wed, 31 Dec 2025 14:16:26 +0000 + DiffCoT: Diffusion-styled Chain-of-Thought Reasoning in LLMs + https://arxiv.org/abs/2601.03559 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03559.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shidong Cao, Hongzhan Lin, Yuxuan Gu, Ziyang Luo, Jing Ma</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Chain-of-Thought (CoT) reasoning improves multi-step mathematical problem solving in large language models but remains vulnerable to exposure bias and error accumulation, as early mistakes propagate irreversibly through autoregressive decoding. In this work, we propose DiffCoT, a diffusion-styled CoT framework that reformulates CoT reasoning as an iterative denoising process. DiffCoT integrates diffusion principles at the reasoning-step level via a sliding-window mechanism, enabling unified generation and retrospective correction of intermediate steps while preserving token-level autoregression. To maintain causal consistency, we further introduce a causal diffusion noise schedule that respects the temporal structure of reasoning chains. Extensive experiments on three multi-step CoT reasoning benchmarks across diverse model backbones demonstrate that DiffCoT consistently outperforms existing CoT preference optimization methods, yielding improved robustness and error-correction capability in CoT reasoning.</p> + https://arxiv.org/abs/2601.03559 + Wed, 07 Jan 2026 03:58:42 +0000 - BEDA: Belief Estimation as Probabilistic Constraints for Performing Strategic Dialogue Acts - https://arxiv.org/abs/2512.24885 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.24885.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hengli Li, Zhaoxin Yu, Qi Shen, Chenxi Li, Mengmeng Wang, Tinglang Wu, Yipeng Kang, Yuxuan Wang, Song-Chun Zhu, Zixia Jia, Zilong Zheng</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Strategic dialogue requires agents to execute distinct dialogue acts, for which belief estimation is essential. While prior work often estimates beliefs accurately, it lacks a principled mechanism to use those beliefs during generation. We bridge this gap by first formalizing two core acts Adversarial and Alignment, and by operationalizing them via probabilistic constraints on what an agent may generate. We instantiate this idea in BEDA, a framework that consists of the world set, the belief estimator for belief estimation, and the conditional generator that selects acts and realizes utterances consistent with the inferred beliefs. Across three settings, Conditional Keeper Burglar (CKBG, adversarial), Mutual Friends (MF, cooperative), and CaSiNo (negotiation), BEDA consistently outperforms strong baselines: on CKBG it improves success rate by at least 5.0 points across backbones and by 20.6 points with GPT-4.1-nano; on Mutual Friends it achieves an average improvement of 9.3 points; and on CaSiNo it achieves the optimal deal relative to all baselines. These results indicate that casting belief estimation as constraints provides a simple, general mechanism for reliable strategic dialogue.</p> - https://arxiv.org/abs/2512.24885 - Wed, 31 Dec 2025 14:26:55 +0000 + RedBench: A Universal Dataset for Comprehensive Red Teaming of Large Language Models + https://arxiv.org/abs/2601.03699 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03699.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Quy-Anh Dang, Chris Ngo, Truong-Son Hy</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> As large language models (LLMs) become integral to safety-critical applications, ensuring their robustness against adversarial prompts is paramount. However, existing red teaming datasets suffer from inconsistent risk categorizations, limited domain coverage, and outdated evaluations, hindering systematic vulnerability assessments. To address these challenges, we introduce RedBench, a universal dataset aggregating 37 benchmark datasets from leading conferences and repositories, comprising 29,362 samples across attack and refusal prompts. RedBench employs a standardized taxonomy with 22 risk categories and 19 domains, enabling consistent and comprehensive evaluations of LLM vulnerabilities. We provide a detailed analysis of existing datasets, establish baselines for modern LLMs, and open-source the dataset and evaluation code. Our contributions facilitate robust comparisons, foster future research, and promote the development of secure and reliable LLMs for real-world deployment. Code: https://github.com/knoveleng/redeval</p> + https://arxiv.org/abs/2601.03699 + Wed, 07 Jan 2026 08:34:17 +0000 - Scaling Open-Ended Reasoning to Predict the Future - https://arxiv.org/abs/2512.25070 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.25070.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Nikhil Chandak, Shashwat Goel, Ameya Prabhu, Moritz Hardt, Jonas Geiping</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> High-stakes decision making involves reasoning under uncertainty about the future. In this work, we train language models to make predictions on open-ended forecasting questions. To scale up training data, we synthesize novel forecasting questions from global events reported in daily news, using a fully automated, careful curation recipe. We train the Qwen3 thinking models on our dataset, OpenForesight. To prevent leakage of future information during training and evaluation, we use an offline news corpus, both for data generation and retrieval in our forecasting system. Guided by a small validation set, we show the benefits of retrieval, and an improved reward function for reinforcement learning (RL). Once we obtain our final forecasting system, we perform held-out testing between May to August 2025. Our specialized model, OpenForecaster 8B, matches much larger proprietary models, with our training improving the accuracy, calibration, and consistency of predictions. We find calibration improvements from forecasting training generalize across popular benchmarks. We open-source all our models, code, and data to make research on language model forecasting broadly accessible.</p> - https://arxiv.org/abs/2512.25070 - Wed, 31 Dec 2025 18:59:51 +0000 + ROI-Reasoning: Rational Optimization for Inference via Pre-Computation Meta-Cognition + https://arxiv.org/abs/2601.03822 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03822.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Muyang Zhao, Qi Qi, Hao Sun</p><p><b>Upvotes:</b> 21</p><p><b>Summary:</b> Large language models (LLMs) can achieve strong reasoning performance with sufficient computation, but they do not inherently know how much computation a task requires. We study budgeted inference-time reasoning for multiple tasks under a strict global token constraint and formalize it as a Ordered Stochastic Multiple-Choice Knapsack Problem(OS-MCKP). This perspective highlights a meta-cognitive requirement -- anticipating task difficulty, estimating return over investment (ROI), and allocating computation strategically. We propose ROI-Reasoning, a two-stage framework that endows LLMs with intrinsic, budget-aware rationality. In the first stage, Meta-Cognitive Fine-Tuning teaches models to predict reasoning cost and expected utility before generation, enabling explicit solve-or-skip decisions. Next, Rationality-Aware Reinforcement Learning optimizes sequential decision making under a hard token budget, allowing models to learn long-horizon allocation strategies. Across budgeted mathematical reasoning benchmarks, ROI-Reasoning consistently improves overall score while substantially reducing regret under tight computation budgets.</p> + https://arxiv.org/abs/2601.03822 + Wed, 07 Jan 2026 11:30:55 +0000 - GaMO: Geometry-aware Multi-view Diffusion Outpainting for Sparse-View 3D Reconstruction - https://arxiv.org/abs/2512.25073 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.25073.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yi-Chuan Huang, Hao-Jen Chien, Chin-Yang Lin, Ying-Huan Chen, Yu-Lun Liu</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> Recent advances in 3D reconstruction have achieved remarkable progress in high-quality scene capture from dense multi-view imagery, yet struggle when input views are limited. Various approaches, including regularization techniques, semantic priors, and geometric constraints, have been implemented to address this challenge. Latest diffusion-based methods have demonstrated substantial improvements by generating novel views from new camera poses to augment training data, surpassing earlier regularization and prior-based techniques. Despite this progress, we identify three critical limitations in these state-of-the-art approaches: inadequate coverage beyond known view peripheries, geometric inconsistencies across generated views, and computationally expensive pipelines. We introduce GaMO (Geometry-aware Multi-view Outpainter), a framework that reformulates sparse-view reconstruction through multi-view outpainting. Instead of generating new viewpoints, GaMO expands the field of view from existing camera poses, which inherently preserves geometric consistency while providing broader scene coverage. Our approach employs multi-view conditioning and geometry-aware denoising strategies in a zero-shot manner without training. Extensive experiments on Replica and ScanNet++ demonstrate state-of-the-art reconstruction quality across 3, 6, and 9 input views, outperforming prior methods in PSNR and LPIPS, while achieving a 25times speedup over SOTA diffusion-based methods with processing time under 10 minutes. Project page: https://yichuanh.github.io/GaMO/</p> - https://arxiv.org/abs/2512.25073 - Wed, 31 Dec 2025 18:59:55 +0000 + Atlas: Orchestrating Heterogeneous Models and Tools for Multi-Domain Complex Reasoning + https://arxiv.org/abs/2601.03872 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03872.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jinyang Wu, Guocheng Zhai, Ruihan Jin, Jiahao Yuan, Yuhao Shen, Shuai Zhang, Zhengqi Wen, Jianhua Tao</p><p><b>Upvotes:</b> 36</p><p><b>Summary:</b> The integration of large language models (LLMs) with external tools has significantly expanded the capabilities of AI agents. However, as the diversity of both LLMs and tools increases, selecting the optimal model-tool combination becomes a high-dimensional optimization challenge. Existing approaches often rely on a single model or fixed tool-calling logic, failing to exploit the performance variations across heterogeneous model-tool pairs. In this paper, we present ATLAS (Adaptive Tool-LLM Alignment and Synergistic Invocation), a dual-path framework for dynamic tool usage in cross-domain complex reasoning. ATLAS operates via a dual-path approach: (1) training-free cluster-based routing that exploits empirical priors for domain-specific alignment, and (2) RL-based multi-step routing that explores autonomous trajectories for out-of-distribution generalization. Extensive experiments across 15 benchmarks demonstrate that our method outperforms closed-source models like GPT-4o, surpassing existing routing methods on both in-distribution (+10.1%) and out-of-distribution (+13.1%) tasks. Furthermore, our framework shows significant gains in visual reasoning by orchestrating specialized multi-modal tools.</p> + https://arxiv.org/abs/2601.03872 + Wed, 07 Jan 2026 12:38:33 +0000 - SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time - https://arxiv.org/abs/2512.25075 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.25075.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhening Huang, Hyeonho Jeong, Xuelin Chen, Yulia Gryaditskaya, Tuanfeng Y. Wang, Joan Lasenby, Chun-Hao Huang</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> We present SpaceTimePilot, a video diffusion model that disentangles space and time for controllable generative rendering. Given a monocular video, SpaceTimePilot can independently alter the camera viewpoint and the motion sequence within the generative process, re-rendering the scene for continuous and arbitrary exploration across space and time. To achieve this, we introduce an effective animation time-embedding mechanism in the diffusion process, allowing explicit control of the output video's motion sequence with respect to that of the source video. As no datasets provide paired videos of the same dynamic scene with continuous temporal variations, we propose a simple yet effective temporal-warping training scheme that repurposes existing multi-view datasets to mimic temporal differences. This strategy effectively supervises the model to learn temporal control and achieve robust space-time disentanglement. To further enhance the precision of dual control, we introduce two additional components: an improved camera-conditioning mechanism that allows altering the camera from the first frame, and CamxTime, the first synthetic space-and-time full-coverage rendering dataset that provides fully free space-time video trajectories within a scene. Joint training on the temporal-warping scheme and the CamxTime dataset yields more precise temporal control. We evaluate SpaceTimePilot on both real-world and synthetic data, demonstrating clear space-time disentanglement and strong results compared to prior work. Project page: https://zheninghuang.github.io/Space-Time-Pilot/ Code: https://github.com/ZheningHuang/spacetimepilot</p> - https://arxiv.org/abs/2512.25075 - Wed, 31 Dec 2025 18:59:57 +0000 + ResTok: Learning Hierarchical Residuals in 1D Visual Tokenizers for Autoregressive Image Generation + https://arxiv.org/abs/2601.03955 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03955.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xu Zhang, Cheng Da, Huan Yang, Kun Gai, Ming Lu, Zhan Ma</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Existing 1D visual tokenizers for autoregressive (AR) generation largely follow the design principles of language modeling, as they are built directly upon transformers whose priors originate in language, yielding single-hierarchy latent tokens and treating visual data as flat sequential token streams. However, this language-like formulation overlooks key properties of vision, particularly the hierarchical and residual network designs that have long been essential for convergence and efficiency in visual models. To bring "vision" back to vision, we propose the Residual Tokenizer (ResTok), a 1D visual tokenizer that builds hierarchical residuals for both image tokens and latent tokens. The hierarchical representations obtained through progressively merging enable cross-level feature fusion at each layer, substantially enhancing representational capacity. Meanwhile, the semantic residuals between hierarchies prevent information overlap, yielding more concentrated latent distributions that are easier for AR modeling. Cross-level bindings consequently emerge without any explicit constraints. To accelerate the generation process, we further introduce a hierarchical AR generator that substantially reduces sampling steps by predicting an entire level of latent tokens at once rather than generating them strictly token-by-token. Extensive experiments demonstrate that restoring hierarchical residual priors in visual tokenization significantly improves AR image generation, achieving a gFID of 2.34 on ImageNet-256 with only 9 sampling steps. Code is available at https://github.com/Kwai-Kolors/ResTok.</p> + https://arxiv.org/abs/2601.03955 + Wed, 07 Jan 2026 14:09:18 +0000 + + + Benchmark^2: Systematic Evaluation of LLM Benchmarks + https://arxiv.org/abs/2601.03986 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.03986.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qi Qian, Chengsong Huang, Jingwen Xu, Changze Lv, Muling Wu, Wenhao Liu, Xiaohua Wang, Zhenghua Wang, Zisu Huang, Muzhao Tian, Jianhan Xu, Kun Hu, He-Da Wang, Yao Hu, Xuanjing Huang, Xiaoqing Zheng</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> The rapid proliferation of benchmarks for evaluating large language models (LLMs) has created an urgent need for systematic methods to assess benchmark quality itself. We propose Benchmark^2, a comprehensive framework comprising three complementary metrics: (1) Cross-Benchmark Ranking Consistency, measuring whether a benchmark produces model rankings aligned with peer benchmarks; (2) Discriminability Score, quantifying a benchmark's ability to differentiate between models; and (3) Capability Alignment Deviation, identifying problematic instances where stronger models fail but weaker models succeed within the same model family. We conduct extensive experiments across 15 benchmarks spanning mathematics, reasoning, and knowledge domains, evaluating 11 LLMs across four model families. Our analysis reveals significant quality variations among existing benchmarks and demonstrates that selective benchmark construction based on our metrics can achieve comparable evaluation performance with substantially reduced test sets.</p> + https://arxiv.org/abs/2601.03986 + Wed, 07 Jan 2026 14:59:03 +0000 + + + Klear: Unified Multi-Task Audio-Video Joint Generation + https://arxiv.org/abs/2601.04151 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04151.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jun Wang, Chunyu Qiang, Yuxin Guo, Yiran Wang, Xijuan Zeng, Chen Zhang, Pengfei Wan</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Audio-video joint generation has progressed rapidly, yet substantial challenges still remain. Non-commercial approaches still suffer audio-visual asynchrony, poor lip-speech alignment, and unimodal degradation, which can be stemmed from weak audio-visual correspondence modeling, limited generalization, and scarce high-quality dense-caption data. To address these issues, we introduce Klear and delve into three axes--model architecture, training strategy, and data curation. Architecturally, we adopt a single-tower design with unified DiT blocks and an Omni-Full Attention mechanism, achieving tight audio-visual alignment and strong scalability. Training-wise, we adopt a progressive multitask regime--random modality masking to joint optimization across tasks, and a multistage curriculum, yielding robust representations, strengthening A-V aligned world knowledge, and preventing unimodal collapse. For datasets, we present the first large-scale audio-video dataset with dense captions, and introduce a novel automated data-construction pipeline which annotates and filters millions of diverse, high-quality, strictly aligned audio-video-caption triplets. Building on this, Klear scales to large datasets, delivering high-fidelity, semantically and temporally aligned, instruction-following generation in both joint and unimodal settings while generalizing robustly to out-of-distribution scenarios. Across tasks, it substantially outperforms prior methods by a large margin and achieves performance comparable to Veo 3, offering a unified, scalable path toward next-generation audio-video synthesis.</p> + https://arxiv.org/abs/2601.04151 + Wed, 07 Jan 2026 18:03:45 +0000 + + + Beyond Binary Preference: Aligning Diffusion Models to Fine-grained Criteria by Decoupling Attributes + https://arxiv.org/abs/2601.04300 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04300.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenye Meng, Zejian Li, Zhongni Liu, Yize Li, Changle Xie, Kaixin Jia, Ling Yang, Huanghuang Deng, Shiying Ding, Shengyuan Zhang, Jiayi Li, Lingyun Sun</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Post-training alignment of diffusion models relies on simplified signals, such as scalar rewards or binary preferences. This limits alignment with complex human expertise, which is hierarchical and fine-grained. To address this, we first construct a hierarchical, fine-grained evaluation criteria with domain experts, which decomposes image quality into multiple positive and negative attributes organized in a tree structure. Building on this, we propose a two-stage alignment framework. First, we inject domain knowledge to an auxiliary diffusion model via Supervised Fine-Tuning. Second, we introduce Complex Preference Optimization (CPO) that extends DPO to align the target diffusion to our non-binary, hierarchical criteria. Specifically, we reformulate the alignment problem to simultaneously maximize the probability of positive attributes while minimizing the probability of negative attributes with the auxiliary diffusion. We instantiate our approach in the domain of painting generation and conduct CPO training with an annotated dataset of painting with fine-grained attributes based on our criteria. Extensive experiments demonstrate that CPO significantly enhances generation quality and alignment with expertise, opening new avenues for fine-grained criteria alignment.</p> + https://arxiv.org/abs/2601.04300 + Wed, 07 Jan 2026 18:11:22 +0000 + + + Agentic Rubrics as Contextual Verifiers for SWE Agents + https://arxiv.org/abs/2601.04171 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04171.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mohit Raghavendra, Anisha Gunjal, Bing Liu, Yunzhong He</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Verification is critical for improving agents: it provides the reward signal for Reinforcement Learning and enables inference-time gains through Test-Time Scaling (TTS). Despite its importance, verification in software engineering (SWE) agent settings often relies on code execution, which can be difficult to scale due to environment setup overhead. Scalable alternatives such as patch classifiers and heuristic methods exist, but they are less grounded in codebase context and harder to interpret. To this end, we explore Agentic Rubrics: an expert agent interacts with the repository to create a context-grounded rubric checklist, and candidate patches are then scored against it without requiring test execution. On SWE-Bench Verified under parallel TTS evaluation, Agentic Rubrics achieve a score of 54.2% on Qwen3-Coder-30B-A3B and 40.6% on Qwen3-32B, with at least a +3.5 percentage-point gain over the strongest baseline in our comparison set. We further analyze rubric behavior, showing that rubric scores are consistent with ground-truth tests while also flagging issues that tests do not capture. Our ablations show that agentic context gathering is essential for producing codebase-specific, unambiguous criteria. Together, these results suggest that Agentic Rubrics provide an efficient, scalable, and granular verification signal for SWE agents.</p> + https://arxiv.org/abs/2601.04171 + Wed, 07 Jan 2026 18:38:23 +0000 + + + Choreographing a World of Dynamic Objects + https://arxiv.org/abs/2601.04194 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04194.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yanzhe Lyu, Chen Geng, Karthik Dharmarajan, Yunzhi Zhang, Hadi Alzayer, Shangzhe Wu, Jiajun Wu</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Dynamic objects in our physical 4D (3D + time) world are constantly evolving, deforming, and interacting with other objects, leading to diverse 4D scene dynamics. In this paper, we present a universal generative pipeline, CHORD, for CHOReographing Dynamic objects and scenes and synthesizing this type of phenomena. Traditional rule-based graphics pipelines to create these dynamics are based on category-specific heuristics, yet are labor-intensive and not scalable. Recent learning-based methods typically demand large-scale datasets, which may not cover all object categories in interest. Our approach instead inherits the universality from the video generative models by proposing a distillation-based pipeline to extract the rich Lagrangian motion information hidden in the Eulerian representations of 2D videos. Our method is universal, versatile, and category-agnostic. We demonstrate its effectiveness by conducting experiments to generate a diverse range of multi-body 4D dynamics, show its advantage compared to existing methods, and demonstrate its applicability in generating robotics manipulation policies. Project page: https://yanzhelyu.github.io/chord</p> + https://arxiv.org/abs/2601.04194 + Wed, 07 Jan 2026 18:59:40 +0000 + + + ReHyAt: Recurrent Hybrid Attention for Video Diffusion Transformers + https://arxiv.org/abs/2601.04342 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04342.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mohsen Ghafoorian, Amirhossein Habibian</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Recent advances in video diffusion models have shifted towards transformer-based architectures, achieving state-of-the-art video generation but at the cost of quadratic attention complexity, which severely limits scalability for longer sequences. We introduce ReHyAt, a Recurrent Hybrid Attention mechanism that combines the fidelity of softmax attention with the efficiency of linear attention, enabling chunk-wise recurrent reformulation and constant memory usage. Unlike the concurrent linear-only SANA Video, ReHyAt's hybrid design allows efficient distillation from existing softmax-based models, reducing the training cost by two orders of magnitude to ~160 GPU hours, while being competitive in the quality. Our light-weight distillation and finetuning pipeline provides a recipe that can be applied to future state-of-the-art bidirectional softmax-based models. Experiments on VBench and VBench-2.0, as well as a human preference study, demonstrate that ReHyAt achieves state-of-the-art video quality while reducing attention cost from quadratic to linear, unlocking practical scalability for long-duration and on-device video generation. Project page is available at https://qualcomm-ai-research.github.io/rehyat.</p> + https://arxiv.org/abs/2601.04342 + Wed, 07 Jan 2026 19:26:30 +0000 + + + Scaling Behavior Cloning Improves Causal Reasoning: An Open Model for Real-Time Video Game Playing + https://arxiv.org/abs/2601.04575 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04575.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuguang Yue, Irakli Salia, Samuel Hunt, Chris Green, Wenzhe Shi, Jonathan J Hunt</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Behavior cloning is enjoying a resurgence in popularity as scaling both model and data sizes proves to provide a strong starting point for many tasks of interest. In this work, we introduce an open recipe for training a video game playing foundation model designed for inference in realtime on a consumer GPU. We release all data (8300+ hours of high quality human gameplay), training and inference code, and pretrained checkpoints under an open license. We show that our best model is capable of playing a variety of 3D video games at a level competitive with human play. We use this recipe to systematically examine the scaling laws of behavior cloning to understand how the model's performance and causal reasoning varies with model and data scale. We first show in a simple toy problem that, for some types of causal reasoning, increasing both the amount of training data and the depth of the network results in the model learning a more causal policy. We then systematically study how causality varies with the number of parameters (and depth) and training steps in scaled models of up to 1.2 billion parameters, and we find similar scaling results to what we observe in the toy problem.</p> + https://arxiv.org/abs/2601.04575 + Thu, 08 Jan 2026 04:06:17 +0000 + + + AgentDevel: Reframing Self-Evolving LLM Agents as Release Engineering + https://arxiv.org/abs/2601.04620 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04620.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Di Zhang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Recent progress in large language model (LLM) agents has largely focused on embedding self-improvement mechanisms inside the agent or searching over many concurrent variants. While these approaches can raise aggregate scores, they often yield unstable and hard-to-audit improvement trajectories, making it difficult to guarantee non-regression or to reason about failures across versions. We reframe agent improvement as release engineering: agents are treated as shippable artifacts, and improvement is externalized into a regression-aware release pipeline. We introduce AgentDevel, a release engineering pipeline that iteratively runs the current agent, produces implementation-blind, symptom-level quality signals from execution traces, synthesizes a single release candidate (RC) via executable diagnosis, and promotes it under flip-centered gating. AgentDevel features three core designs: (i) an implementation-blind LLM critic that characterizes failure appearances without accessing agent internals, (ii) script-based executable diagnosis that aggregates dominant symptom patterns and produces auditable engineering specifications, and (iii) flip-centered gating that prioritizes pass to fail regressions and fail to pass fixes as first-class evidence. Unlike population-based search or in-agent self-refinement, AgentDevel maintains a single canonical version line and emphasizes non-regression as a primary objective. Experiments on execution-heavy benchmarks demonstrate that AgentDevel yields stable improvements with significantly fewer regressions while producing reproducible, auditable artifacts. Overall, AgentDevel provides a practical development discipline for building, debugging, and releasing LLM agents as software development.</p> + https://arxiv.org/abs/2601.04620 + Thu, 08 Jan 2026 05:49:01 +0000 + + + ProFuse: Efficient Cross-View Context Fusion for Open-Vocabulary 3D Gaussian Splatting + https://arxiv.org/abs/2601.04754 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04754.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yen-Jen Chiou, Wei-Tse Cheng, Yuan-Fu Yang</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> We present ProFuse, an efficient context-aware framework for open-vocabulary 3D scene understanding with 3D Gaussian Splatting (3DGS). The pipeline enhances cross-view consistency and intra-mask cohesion within a direct registration setup, adding minimal overhead and requiring no render-supervised fine-tuning. Instead of relying on a pretrained 3DGS scene, we introduce a dense correspondence-guided pre-registration phase that initializes Gaussians with accurate geometry while jointly constructing 3D Context Proposals via cross-view clustering. Each proposal carries a global feature obtained through weighted aggregation of member embeddings, and this feature is fused onto Gaussians during direct registration to maintain per-primitive language coherence across views. With associations established in advance, semantic fusion requires no additional optimization beyond standard reconstruction, and the model retains geometric refinement without densification. ProFuse achieves strong open-vocabulary 3DGS understanding while completing semantic attachment in about five minutes per scene, which is two times faster than SOTA.</p> + https://arxiv.org/abs/2601.04754 + Thu, 08 Jan 2026 09:20:46 +0000 + + + AT^2PO: Agentic Turn-based Policy Optimization via Tree Search + https://arxiv.org/abs/2601.04767 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04767.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zefang Zong, Dingwei Chen, Yang Li, Qi Yi, Bo Zhou, Chengming Li, Bo Qian, Peng Chen, Jie Jiang</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> LLM agents have emerged as powerful systems for tackling multi-turn tasks by interleaving internal reasoning and external tool interactions. Agentic Reinforcement Learning has recently drawn significant research attention as a critical post-training paradigm to further refine these capabilities. In this paper, we present AT^2PO (Agentic Turn-based Policy Optimization via Tree Search), a unified framework for multi-turn agentic RL that addresses three core challenges: limited exploration diversity, sparse credit assignment, and misaligned policy optimization. AT^2PO introduces a turn-level tree structure that jointly enables Entropy-Guided Tree Expansion for strategic exploration and Turn-wise Credit Assignment for fine-grained reward propagation from sparse outcomes. Complementing this, we propose Agentic Turn-based Policy Optimization, a turn-level learning objective that aligns policy updates with the natural decision granularity of agentic interactions. ATPO is orthogonal to tree search and can be readily integrated into any multi-turn RL pipeline. Experiments across seven benchmarks demonstrate consistent improvements over the state-of-the-art baseline by up to 1.84 percentage points in average, with ablation studies validating the effectiveness of each component. Our code is available at https://github.com/zzfoutofspace/ATPO.</p> + https://arxiv.org/abs/2601.04767 + Thu, 08 Jan 2026 09:35:49 +0000 + + + PyramidalWan: On Making Pretrained Video Model Pyramidal for Efficient Inference + https://arxiv.org/abs/2601.04792 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04792.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Denis Korzhenkov, Adil Karjauv, Animesh Karnewar, Mohsen Ghafoorian, Amirhossein Habibian</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Recently proposed pyramidal models decompose the conventional forward and backward diffusion processes into multiple stages operating at varying resolutions. These models handle inputs with higher noise levels at lower resolutions, while less noisy inputs are processed at higher resolutions. This hierarchical approach significantly reduces the computational cost of inference in multi-step denoising models. However, existing open-source pyramidal video models have been trained from scratch and tend to underperform compared to state-of-the-art systems in terms of visual plausibility. In this work, we present a pipeline that converts a pretrained diffusion model into a pyramidal one through low-cost finetuning, achieving this transformation without degradation in quality of output videos. Furthermore, we investigate and compare various strategies for step distillation within pyramidal models, aiming to further enhance the inference efficiency. Our results are available at https://qualcomm-ai-research.github.io/PyramidalWan.</p> + https://arxiv.org/abs/2601.04792 + Thu, 08 Jan 2026 10:16:06 +0000 + + + Learnable Multipliers: Freeing the Scale of Language Model Matrix Layers + https://arxiv.org/abs/2601.04890 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.04890.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Maksim Velikanov, Ilyas Chahed, Jingwei Zuo, Dhia Eddine Rhaiem, Younes Belkada, Hakim Hacid</p><p><b>Upvotes:</b> 36</p><p><b>Summary:</b> Applying weight decay (WD) to matrix layers is standard practice in large-language-model pretraining. Prior work suggests that stochastic gradient noise induces a Brownian-like expansion of the weight matrices W, whose growth is counteracted by WD, leading to a WD-noise equilibrium with a certain weight norm ||W||. In this work, we view the equilibrium norm as a harmful artifact of the training procedure, and address it by introducing learnable multipliers to learn the optimal scale. First, we attach a learnable scalar multiplier to W and confirm that the WD-noise equilibrium norm is suboptimal: the learned scale adapts to data and improves performance. We then argue that individual row and column norms are similarly constrained, and free their scale by introducing learnable per-row and per-column multipliers. Our method can be viewed as a learnable, more expressive generalization of muP multipliers. It outperforms a well-tuned muP baseline, reduces the computational overhead of multiplier tuning, and surfaces practical questions such as forward-pass symmetries and the width-scaling of the learned multipliers. Finally, we validate learnable multipliers with both Adam and Muon optimizers, where it shows improvement in downstream evaluations matching the improvement of the switching from Adam to Muon.</p> + https://arxiv.org/abs/2601.04890 + Thu, 08 Jan 2026 12:41:49 +0000 + + + Token-Level LLM Collaboration via FusionRoute + https://arxiv.org/abs/2601.05106 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05106.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Nuoya Xiong, Yuhang Zhou, Hanqing Zeng, Zhaorun Chen, Furong Huang, Shuchao Bi, Lizhu Zhang, Zhuokai Zhao</p><p><b>Upvotes:</b> 30</p><p><b>Summary:</b> Large language models (LLMs) exhibit strengths across diverse domains. However, achieving strong performance across these domains with a single general-purpose model typically requires scaling to sizes that are prohibitively expensive to train and deploy. On the other hand, while smaller domain-specialized models are much more efficient, they struggle to generalize beyond their training distributions. To address this dilemma, we propose FusionRoute, a robust and effective token-level multi-LLM collaboration framework in which a lightweight router simultaneously (i) selects the most suitable expert at each decoding step and (ii) contributes a complementary logit that refines or corrects the selected expert's next-token distribution via logit addition. Unlike existing token-level collaboration methods that rely solely on fixed expert outputs, we provide a theoretical analysis showing that pure expert-only routing is fundamentally limited: unless strong global coverage assumptions hold, it cannot in general realize the optimal decoding policy. By augmenting expert selection with a trainable complementary generator, FusionRoute expands the effective policy class and enables recovery of optimal value functions under mild conditions. Empirically, across both Llama-3 and Gemma-2 families and diverse benchmarks spanning mathematical reasoning, code generation, and instruction following, FusionRoute outperforms both sequence- and token-level collaboration, model merging, and direct fine-tuning, while remaining competitive with domain experts on their respective tasks.</p> + https://arxiv.org/abs/2601.05106 + Thu, 08 Jan 2026 16:53:16 +0000 + + + Agent-as-a-Judge + https://arxiv.org/abs/2601.05111 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05111.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Runyang You, Hongru Cai, Caiqi Zhang, Qiancheng Xu, Meng Liu, Tiezheng Yu, Yongqi Li, Wenjie Li</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> LLM-as-a-Judge has revolutionized AI evaluation by leveraging large language models for scalable assessments. However, as evaluands become increasingly complex, specialized, and multi-step, the reliability of LLM-as-a-Judge has become constrained by inherent biases, shallow single-pass reasoning, and the inability to verify assessments against real-world observations. This has catalyzed the transition to Agent-as-a-Judge, where agentic judges employ planning, tool-augmented verification, multi-agent collaboration, and persistent memory to enable more robust, verifiable, and nuanced evaluations. Despite the rapid proliferation of agentic evaluation systems, the field lacks a unified framework to navigate this shifting landscape. To bridge this gap, we present the first comprehensive survey tracing this evolution. Specifically, we identify key dimensions that characterize this paradigm shift and establish a developmental taxonomy. We organize core methodologies and survey applications across general and professional domains. Furthermore, we analyze frontier challenges and identify promising research directions, ultimately providing a clear roadmap for the next generation of agentic evaluation.</p> + https://arxiv.org/abs/2601.05111 + Thu, 08 Jan 2026 16:58:10 +0000 + + + Re-Align: Structured Reasoning-guided Alignment for In-Context Image Generation and Editing + https://arxiv.org/abs/2601.05124 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05124.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Runze He, Yiji Cheng, Tiankai Hang, Zhimin Li, Yu Xu, Zijin Yin, Shiyi Zhang, Wenxun Dai, Penghui Du, Ao Ma, Chunyu Wang, Qinglin Lu, Jizhong Han, Jiao Dai</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> In-context image generation and editing (ICGE) enables users to specify visual concepts through interleaved image-text prompts, demanding precise understanding and faithful execution of user intent. Although recent unified multimodal models exhibit promising understanding capabilities, these strengths often fail to transfer effectively to image generation. We introduce Re-Align, a unified framework that bridges the gap between understanding and generation through structured reasoning-guided alignment. At its core lies the In-Context Chain-of-Thought (IC-CoT), a structured reasoning paradigm that decouples semantic guidance and reference association, providing clear textual target and mitigating confusion among reference images. Furthermore, Re-Align introduces an effective RL training scheme that leverages a surrogate reward to measure the alignment between structured reasoning text and the generated image, thereby improving the model's overall performance on ICGE tasks. Extensive experiments verify that Re-Align outperforms competitive methods of comparable model scale and resources on both in-context image generation and editing tasks.</p> + https://arxiv.org/abs/2601.05124 + Thu, 08 Jan 2026 17:13:00 +0000 + + + VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control + https://arxiv.org/abs/2601.05138 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05138.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Sixiao Zheng, Minghao Yin, Wenbo Hu, Xiaoyu Li, Ying Shan, Yanwei Fu</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Video world models aim to simulate dynamic, real-world environments, yet existing methods struggle to provide unified and precise control over camera and multi-object motion, as videos inherently operate dynamics in the projected 2D image plane. To bridge this gap, we introduce VerseCrafter, a 4D-aware video world model that enables explicit and coherent control over both camera and object dynamics within a unified 4D geometric world state. Our approach is centered on a novel 4D Geometric Control representation, which encodes the world state through a static background point cloud and per-object 3D Gaussian trajectories. This representation captures not only an object's path but also its probabilistic 3D occupancy over time, offering a flexible, category-agnostic alternative to rigid bounding boxes or parametric models. These 4D controls are rendered into conditioning signals for a pretrained video diffusion model, enabling the generation of high-fidelity, view-consistent videos that precisely adhere to the specified dynamics. Unfortunately, another major challenge lies in the scarcity of large-scale training data with explicit 4D annotations. We address this by developing an automatic data engine that extracts the required 4D controls from in-the-wild videos, allowing us to train our model on a massive and diverse dataset.</p> + https://arxiv.org/abs/2601.05138 + Thu, 08 Jan 2026 17:28:52 +0000 + + + Multi-Scale Local Speculative Decoding for Image Generation + https://arxiv.org/abs/2601.05149 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05149.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Elia Peruzzo, Guillaume Sautière, Amirhossein Habibian</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Autoregressive (AR) models have achieved remarkable success in image synthesis, yet their sequential nature imposes significant latency constraints. Speculative Decoding offers a promising avenue for acceleration, but existing approaches are limited by token-level ambiguity and lack of spatial awareness. In this work, we introduce Multi-Scale Local Speculative Decoding (MuLo-SD), a novel framework that combines multi-resolution drafting with spatially informed verification to accelerate AR image generation. Our method leverages a low-resolution drafter paired with learned up-samplers to propose candidate image tokens, which are then verified in parallel by a high-resolution target model. Crucially, we incorporate a local rejection and resampling mechanism, enabling efficient correction of draft errors by focusing on spatial neighborhoods rather than raster-scan resampling after the first rejection. We demonstrate that MuLo-SD achieves substantial speedups - up to 1.7times - outperforming strong speculative decoding baselines such as EAGLE-2 and LANTERN in terms of acceleration, while maintaining comparable semantic alignment and perceptual quality. These results are validated using GenEval, DPG-Bench, and FID/HPSv2 on the MS-COCO 5k validation split. Extensive ablations highlight the impact of up-sampling design, probability pooling, and local rejection and resampling with neighborhood expansion. Our approach sets a new state-of-the-art in speculative decoding for image synthesis, bridging the gap between efficiency and fidelity.</p> + https://arxiv.org/abs/2601.05149 + Thu, 08 Jan 2026 17:39:35 +0000 + + + DocDancer: Towards Agentic Document-Grounded Information Seeking + https://arxiv.org/abs/2601.05163 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05163.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Qintong Zhang, Xinjie Lv, Jialong Wu, Baixuan Li, Zhengwei Tao, Guochen Yan, Huanyao Zhang, Bin Wang, Jiahao Xu, Haitao Mi, Wentao Zhang</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Document Question Answering (DocQA) focuses on answering questions grounded in given documents, yet existing DocQA agents lack effective tool utilization and largely rely on closed-source models. In this work, we introduce DocDancer, an end-to-end trained open-source Doc agent. We formulate DocQA as an information-seeking problem and propose a tool-driven agent framework that explicitly models document exploration and comprehension. To enable end-to-end training of such agents, we introduce an Exploration-then-Synthesis data synthesis pipeline that addresses the scarcity of high-quality training data for DocQA. Training on the synthesized data, the trained models on two long-context document understanding benchmarks, MMLongBench-Doc and DocBench, show their effectiveness. Further analysis provides valuable insights for the agentic tool design and synthetic data.</p> + https://arxiv.org/abs/2601.05163 + Thu, 08 Jan 2026 17:54:32 +0000 + + + RelayLLM: Efficient Reasoning via Collaborative Decoding + https://arxiv.org/abs/2601.05167 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05167.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chengsong Huang, Tong Zheng, Langlin Huang, Jinyuan Li, Haolin Liu, Jiaxin Huang</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> Large Language Models (LLMs) for complex reasoning is often hindered by high computational costs and latency, while resource-efficient Small Language Models (SLMs) typically lack the necessary reasoning capacity. Existing collaborative approaches, such as cascading or routing, operate at a coarse granularity by offloading entire queries to LLMs, resulting in significant computational waste when the SLM is capable of handling the majority of reasoning steps. To address this, we propose RelayLLM, a novel framework for efficient reasoning via token-level collaborative decoding. Unlike routers, RelayLLM empowers the SLM to act as an active controller that dynamically invokes the LLM only for critical tokens via a special command, effectively "relaying" the generation process. We introduce a two-stage training framework, including warm-up and Group Relative Policy Optimization (GRPO) to teach the model to balance independence with strategic help-seeking. Empirical results across six benchmarks demonstrate that RelayLLM achieves an average accuracy of 49.52%, effectively bridging the performance gap between the two models. Notably, this is achieved by invoking the LLM for only 1.07% of the total generated tokens, offering a 98.2% cost reduction compared to performance-matched random routers.</p> + https://arxiv.org/abs/2601.05167 + Thu, 08 Jan 2026 17:56:16 +0000 + + + CoV: Chain-of-View Prompting for Spatial Reasoning + https://arxiv.org/abs/2601.05172 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05172.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoyu Zhao, Akide Liu, Zeyu Zhang, Weijie Wang, Feng Chen, Ruihan Zhu, Gholamreza Haffari, Bohan Zhuang</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Embodied question answering (EQA) in 3D environments often requires collecting context that is distributed across multiple viewpoints and partially occluded. However, most recent vision--language models (VLMs) are constrained to a fixed and finite set of input views, which limits their ability to acquire question-relevant context at inference time and hinders complex spatial reasoning. We propose Chain-of-View (CoV) prompting, a training-free, test-time reasoning framework that transforms a VLM into an active viewpoint reasoner through a coarse-to-fine exploration process. CoV first employs a View Selection agent to filter redundant frames and identify question-aligned anchor views. It then performs fine-grained view adjustment by interleaving iterative reasoning with discrete camera actions, obtaining new observations from the underlying 3D scene representation until sufficient context is gathered or a step budget is reached. We evaluate CoV on OpenEQA across four mainstream VLMs and obtain an average +11.56\% improvement in LLM-Match, with a maximum gain of +13.62\% on Qwen3-VL-Flash. CoV further exhibits test-time scaling: increasing the minimum action budget yields an additional +2.51\% average improvement, peaking at +3.73\% on Gemini-2.5-Flash. On ScanQA and SQA3D, CoV delivers strong performance (e.g., 116 CIDEr / 31.9 EM@1 on ScanQA and 51.1 EM@1 on SQA3D). Overall, these results suggest that question-aligned view selection coupled with open-view search is an effective, model-agnostic strategy for improving spatial reasoning in 3D EQA without additional training.</p> + https://arxiv.org/abs/2601.05172 + Thu, 08 Jan 2026 17:59:42 +0000 + + + VideoAuto-R1: Video Auto Reasoning via Thinking Once, Answering Twice + https://arxiv.org/abs/2601.05175 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05175.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuming Liu, Mingchen Zhuge, Changsheng Zhao, Jun Chen, Lemeng Wu, Zechun Liu, Chenchen Zhu, Zhipeng Cai, Chong Zhou, Haozhe Liu, Ernie Chang, Saksham Suri, Hongyu Xu, Qi Qian, Wei Wen, Balakrishnan Varadarajan, Zhuang Liu, Hu Xu, Florian Bordes, Raghuraman Krishnamoorthi, Bernard Ghanem, Vikas Chandra, Yunyang Xiong</p><p><b>Upvotes:</b> 25</p><p><b>Summary:</b> Chain-of-thought (CoT) reasoning has emerged as a powerful tool for multimodal large language models on video understanding tasks. However, its necessity and advantages over direct answering remain underexplored. In this paper, we first demonstrate that for RL-trained video models, direct answering often matches or even surpasses CoT performance, despite CoT producing step-by-step analyses at a higher computational cost. Motivated by this, we propose VideoAuto-R1, a video understanding framework that adopts a reason-when-necessary strategy. During training, our approach follows a Thinking Once, Answering Twice paradigm: the model first generates an initial answer, then performs reasoning, and finally outputs a reviewed answer. Both answers are supervised via verifiable rewards. During inference, the model uses the confidence score of the initial answer to determine whether to proceed with reasoning. Across video QA and grounding benchmarks, VideoAuto-R1 achieves state-of-the-art accuracy with significantly improved efficiency, reducing the average response length by ~3.3x, e.g., from 149 to just 44 tokens. Moreover, we observe a low rate of thinking-mode activation on perception-oriented tasks, but a higher rate on reasoning-intensive tasks. This suggests that explicit language-based reasoning is generally beneficial but not always necessary.</p> + https://arxiv.org/abs/2601.05175 + Thu, 08 Jan 2026 18:00:59 +0000 + + + Plenoptic Video Generation + https://arxiv.org/abs/2601.05239 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05239.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiao Fu, Shitao Tang, Min Shi, Xian Liu, Jinwei Gu, Ming-Yu Liu, Dahua Lin, Chen-Hsuan Lin</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> Camera-controlled generative video re-rendering methods, such as ReCamMaster, have achieved remarkable progress. However, despite their success in single-view setting, these works often struggle to maintain consistency across multi-view scenarios. Ensuring spatio-temporal coherence in hallucinated regions remains challenging due to the inherent stochasticity of generative models. To address it, we introduce PlenopticDreamer, a framework that synchronizes generative hallucinations to maintain spatio-temporal memory. The core idea is to train a multi-in-single-out video-conditioned model in an autoregressive manner, aided by a camera-guided video retrieval strategy that adaptively selects salient videos from previous generations as conditional inputs. In addition, Our training incorporates progressive context-scaling to improve convergence, self-conditioning to enhance robustness against long-range visual degradation caused by error accumulation, and a long-video conditioning mechanism to support extended video generation. Extensive experiments on the Basic and Agibot benchmarks demonstrate that PlenopticDreamer achieves state-of-the-art video re-rendering, delivering superior view synchronization, high-fidelity visuals, accurate camera control, and diverse view transformations (e.g., third-person to third-person, and head-view to gripper-view in robotic manipulation). Project page: https://research.nvidia.com/labs/dir/plenopticdreamer/</p> + https://arxiv.org/abs/2601.05239 + Thu, 08 Jan 2026 18:58:32 +0000 + + + RoboVIP: Multi-View Video Generation with Visual Identity Prompting Augments Robot Manipulation + https://arxiv.org/abs/2601.05241 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05241.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Boyang Wang, Haoran Zhang, Shujie Zhang, Jinkun Hao, Mingda Jia, Qi Lv, Yucheng Mao, Zhaoyang Lyu, Jia Zeng, Xudong Xu, Jiangmiao Pang</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> The diversity, quantity, and quality of manipulation data are critical for training effective robot policies. However, due to hardware and physical setup constraints, collecting large-scale real-world manipulation data remains difficult to scale across diverse environments. Recent work uses text-prompt conditioned image diffusion models to augment manipulation data by altering the backgrounds and tabletop objects in the visual observations. However, these approaches often overlook the practical need for multi-view and temporally coherent observations required by state-of-the-art policy models. Further, text prompts alone cannot reliably specify the scene setup. To provide the diffusion model with explicit visual guidance, we introduce visual identity prompting, which supplies exemplar images as conditioning inputs to guide the generation of the desired scene setup. To this end, we also build a scalable pipeline to curate a visual identity pool from large robotics datasets. Using our augmented manipulation data to train downstream vision-language-action and visuomotor policy models yields consistent performance gains in both simulation and real-robot settings.</p> + https://arxiv.org/abs/2601.05241 + Thu, 08 Jan 2026 18:59:22 +0000 + + + GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization + https://arxiv.org/abs/2601.05242 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05242.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shih-Yang Liu, Xin Dong, Ximing Lu, Shizhe Diao, Peter Belcak, Mingjie Liu, Min-Hung Chen, Hongxu Yin, Yu-Chiang Frank Wang, Kwang-Ting Cheng, Yejin Choi, Jan Kautz, Pavlo Molchanov</p><p><b>Upvotes:</b> 137</p><p><b>Summary:</b> As language models become increasingly capable, users expect them to provide not only accurate responses but also behaviors aligned with diverse human preferences across a variety of scenarios. To achieve this, Reinforcement learning (RL) pipelines have begun incorporating multiple rewards, each capturing a distinct preference, to guide models toward these desired behaviors. However, recent work has defaulted to apply Group Relative Policy Optimization (GRPO) under multi-reward setting without examining its suitability. In this paper, we demonstrate that directly applying GRPO to normalize distinct rollout reward combinations causes them to collapse into identical advantage values, reducing the resolution of the training signal and resulting in suboptimal convergence and, in some cases, early training failure. We then introduce Group reward-Decoupled Normalization Policy Optimization (GDPO), a new policy optimization method to resolve these issues by decoupling the normalization of individual rewards, more faithfully preserving their relative differences and enabling more accurate multi-reward optimization, along with substantially improved training stability. We compare GDPO with GRPO across three tasks: tool calling, math reasoning, and coding reasoning, evaluating both correctness metrics (accuracy, bug ratio) and constraint adherence metrics (format, length). Across all settings, GDPO consistently outperforms GRPO, demonstrating its effectiveness and generalizability for multi-reward reinforcement learning optimization.</p> + https://arxiv.org/abs/2601.05242 + Thu, 08 Jan 2026 18:59:24 +0000 + + + RL-AWB: Deep Reinforcement Learning for Auto White Balance Correction in Low-Light Night-time Scenes + https://arxiv.org/abs/2601.05249 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2601.05249.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuan-Kang Lee, Kuan-Lin Chen, Chia-Che Chang, Yu-Lun Liu</p><p><b>Upvotes:</b> 35</p><p><b>Summary:</b> Nighttime color constancy remains a challenging problem in computational photography due to low-light noise and complex illumination conditions. We present RL-AWB, a novel framework combining statistical methods with deep reinforcement learning for nighttime white balance. Our method begins with a statistical algorithm tailored for nighttime scenes, integrating salient gray pixel detection with novel illumination estimation. Building on this foundation, we develop the first deep reinforcement learning approach for color constancy that leverages the statistical algorithm as its core, mimicking professional AWB tuning experts by dynamically optimizing parameters for each image. To facilitate cross-sensor evaluation, we introduce the first multi-sensor nighttime dataset. Experiment results demonstrate that our method achieves superior generalization capability across low-light and well-illuminated images. Project page: https://ntuneillee.github.io/research/rl-awb/</p> + https://arxiv.org/abs/2601.05249 + Thu, 08 Jan 2026 18:59:55 +0000