diff --git a/feed.xml b/feed.xml index 73bac56..4c8e9e1 100644 --- a/feed.xml +++ b/feed.xml @@ -7,328 +7,412 @@ http://www.rssboard.org/rss-specification python-feedgen en - Tue, 09 Jun 2026 00:19:54 +0000 + Wed, 10 Jun 2026 00:25:40 +0000 - SPACENUM: Revisiting Spatial Numerical Understanding in VLMs - https://arxiv.org/abs/2605.23898 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.23898.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jianshu Zhang, Yijiang Li, Huifeixin Chen, Haoran Lu, Letian Xue, Bingyang Wang, Han Liu</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Vision-Language Models (VLMs) are increasingly deployed in embodied environments, where they need produce numerical outputs such as action magnitudes and spatial coordinates. Although these numbers appear meaningful, it remains unclear whether these numerical outputs are genuinely grounded in spatial perception. Therefore, in this work, we revisit spatial numerical understanding through SpaceNum, a unified framework that captures two complementary settings: numbers as dynamic transitions during spatial exploration, and numbers as static layouts in spatial reasoning. We formulate two bidirectional tasks, Num2Space and Space2Num, to evaluate how well VLMs map between vision-side spatial structure and language-side numerical representations. We systematically study whether current VLMs truly understand numerical values in spatial settings. Across dynamic transitions and static layouts, we find that models largely fail to ground numbers in spatial meaning and often perform close to random guess. Through error analysis, reasoning trace analysis, and controlled interventions, we show that current VLMs rely heavily on shallow spatial cues, struggle to build stable coordinate-aware representations, and fail to abstract structured spatial layouts from visual observations. We further show that explicit reasoning provides only marginal gains, while tuning can partially improve spatial numerical understanding and transfer to external spatial reasoning benchmarks.</p> - https://arxiv.org/abs/2605.23898 - Fri, 22 May 2026 00:00:00 +0000 + Liberating LLM Capabilities in Full-Duplex Speech Models + https://arxiv.org/abs/2606.07547 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07547.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Luoyuan Zhang, Bokai Xu, Junbo Cui, Weiyue Sun, Yingjing Xu, Hanyu Liu, Yuan Yao</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Speech-based large language models are typically constrained to spoken replies, which limits their user-facing outputs to what can be verbalized and suppresses text-native capabilities such as code generation, structured analysis, and multi-step reasoning in realtime interaction, for tasks that require persistent, structured, and inspectable intermediate outputs. Existing work improves spoken reasoning or full-duplex turn-taking, but still treats text as a hidden intermediate state or a subordinate modality rather than a first-class output channel. We propose Listen-Write-Speak (LWS), a text-first tri-channel paradigm in which a single autoregressive LLM continuously listens to user audio, writes visible free-form text as its primary output, and speaks a realtime oral response in parallel under a shared causal attention context. This behavior is implemented entirely through a Token Schema, requiring no architectural modifications, and learned via a two-stage data pipeline that synthesizes per-second cognitive annotations consistent with the revealed input timeline. Empirically, LWS demonstrates strong full-duplex interaction on Full-Duplex-Bench, reaches 4.72 on VoiceBench AlpacaEval, achieves 92.6% writing-speaking consistency, and consistently outperforms its internal ablations on URO-Bench. These results suggest that visible writing can serve as a first-class output channel for speech interaction without sacrificing realtime responsiveness. The code and dataset are available on the project page: https://royalzhang.com/project/lws-page/.</p> + https://arxiv.org/abs/2606.07547 + Mon, 04 May 2026 00:00:00 +0000 - When Gradients Collide: Failure Modes of Multi-Objective Prompt Optimization for LLM Judges - https://arxiv.org/abs/2605.26046 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.26046.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Parth Darshan, Abhishek Divekar</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Customizing an LLM judge to a specific task or domain often involves optimizing its prompt across multiple evaluation criteria simultaneously. Textual gradient methods automate this for a single judge criterion, however they produce natural-language critiques, not numerical vectors. Thus, the conflict-resolution toolkit of multi-task learning (PCGrad, MGDA) doesn't apply to the multi-objective textual gradient setting. We test five decomposition modes of textual gradient optimizers by varying how much cross-task information the loss, gradient and optimizer LLMs share. In 6 of 10 configurations, we observe that optimization never improves over the initial prompt. Gradient specificity drops by 59% (from 9.0 to 3.7) when the gradient LLM processes multiple criteria jointly. Separately, we observe that naively combining per-task instructions into a single prompt degrades Spearman's rho by -5.3%. These results identify two separable failure modes: optimization-time gradient dilution and inference-time instruction interference, which together constrain the design space for multi-objective judge customization using textual feedback.</p> - https://arxiv.org/abs/2605.26046 - Mon, 25 May 2026 00:00:00 +0000 + EMMA: Extracting Multiple physical parameters from Multimodal Data + https://arxiv.org/abs/2605.24047 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.24047.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Farhat Shaikh, Ayan Banerjee, Sandeep Gupta</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> We introduce EMMA, a physics-informed multimodal framework that recovers all identifiable dynamical parameters of a system directly from raw video, audio, and image-based time-series observations. Unlike prior video-only approaches that struggle with occluded states, hidden actuation inputs, or assumptions about known initial conditions and coordinate frames, EMMA performs joint inference of explicit parameters, implicit dynamical components, and calibration invariants within a unified continuous-time model. EMMA leverages a Liquid Time-Constant (LTC) network to learn latent dynamics from heterogeneous modalities while a physics-constrained loss enforces consistency with the governing differential equations. A unified feature pipeline enables consistent alignment across video trajectories, acoustic signatures, and chart-derived measurements, allowing EMMA to estimate parameters under forced, implicit, and multivariate dynamics without requiring segmentation masks, differentiable rendering, or specialized sensors. Across 100+ scenarios including five standard dynamical benchmarks (75 Delfys videos), real-world rover and quadrotor systems with hidden inputs, and simulation-chart case studies spanning biological and chaotic systems, EMMA delivers robust multi-parameter recovery and significantly outperforms existing single-modality and equation-discovery baselines. Our results establish EMMA as a general, scalable solution for physics-consistent model extraction from opportunistic multimodal data. Code and data are available at: https://github.com/ImpactLabASU/EMMA-CVPR2026</p> + https://arxiv.org/abs/2605.24047 + Thu, 21 May 2026 00:00:00 +0000 - SIA: Self Improving AI with Harness & Weight Updates - https://arxiv.org/abs/2605.27276 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.27276.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Prannay Hebbar, Yogendra Manawat, Samuel Verboomen, Alesia Ivanova, Selvam Palanimalai, Kunal Bhatia, Vignesh Baskaran</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Humans are the bottleneck in building and improving AI. Both the models and the agents that wrap them are written, tuned, and corrected by people. The long-horizon goal of an AI that can figure out how to improve itself remains open. Two largely disjoint research lines attack this bottleneck. The harness-update school has a meta-agent rewrite the scaffold of a task-specific agent (its tools, prompts, retry logic, and search procedure) while the model weights are held fixed. The test-time training school uses hand-written RL pipelines to update the model's own weights on task feedback while the harness is held fixed. These two silos operate in isolation. We propose SIA, a self-improving loop in which a language-model agent (the Feedback-Agent) updates both the harness and the weights of a task-specific agent. We evaluate across three contrasting domains: Chinese legal charge classification, low-level GPU kernel optimisation, and single-cell RNA denoising. Combining both levers outperforms scaffold iteration alone on all three benchmarks. The gains are 56.6% on LawBench, 91.9% runtime reduction on GPU kernels, and 502% on denoising over the initial baseline. Harness updates make the model agentic, shaping how it searches and acts, while weight updates build the domain intuition that no prompt or scaffold can instil.</p> - https://arxiv.org/abs/2605.27276 + WorldCraft: From Camera Navigation to Object Manipulation in Interactive Video World Models + https://arxiv.org/abs/2605.25077 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.25077.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Bohai Gu, Taiyi Wu, Yueyang Yuan, Jian Liu, Xiaocheng Lu, Dazhao Du, Jie Zhang, Jinxiang Lai, Shuai Yang, Xiaotong Zhao, Alan Zhao, Song Guo</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Recent video-based world models have made pixel-space environments interactive at the camera level: users can navigate viewpoints while the model generates coherent visual continuations. Yet their action spaces remain incomplete: users can move the camera, but cannot act on individual objects. Since real-world interaction is inherently object-centric, such models remain closer to passive scene observers than truly manipulable environments. We present WorldCraft, a framework that expands interactive video world models from camera navigation to object-level trajectory actions. Given a user click and a sketched path, WorldCraft generates future frames in which the selected object follows the prescribed trajectory while the camera continues to navigate the scene. WorldCraft achieves this through a trajectory-centric control pipeline: First, Normalized World Trajectory (NWT) represents user-drawn motion in a camera-invariant world coordinate system and dynamically re-projects it under the current camera pose, separating object motion from camera-induced screen-space displacement; Spatial-Pathway LoRA (SP-LoRA) then injects this world-space signal through the model's spatial-control pathway, adding object manipulation capability while preserving the pretrained camera controller; finally, Trajectory-Anchored State Persistence (TASP) treats the world trajectory as a persistent spatial state and refreshes autoregressive memory after trajectory-conditioned generation, allowing moved objects to reappear at their updated positions after leaving the camera view. Experiments show that WorldCraft enables accurate object control, preserves the video-based world model's camera fidelity under camera-only evaluation, and maintains object state across long autoregressive rollouts with off-camera excursions.</p> + https://arxiv.org/abs/2605.25077 + Sun, 24 May 2026 00:00:00 +0000 + + + DEI: Diversity in Evolutionary Inference for Quality-Diversity Search + https://arxiv.org/abs/2605.27130 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.27130.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> John Donaghy, Shikhar Rastogi</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> We present DEI: Diversity in Evolutionary Inference, a distributed Quality-Diversity (QD) search framework that assigns heterogeneous large language models (LLMs) as mutation operators across peer nodes communicating with non-blocking collective operations. Unlike homogeneous parallel search, which replicates a single model's inductive biases across all workers, DEI treats each LLM's distinct creative prior as a complementary source of behavioral novelty. Extending the Digital Red Queen framework with DEI, nodes share local optimal solutions at the end of each round to seed the next round's population. This creates cross-model adversarial pressure that drives robustness beyond intra-model self-play. Evaluated on the Core War domain, a competitive programming benchmark in which Redcode warrior programs battle inside a simulated machine, a four-node heterogeneous ensemble (GPT-5.4-mini, Claude Sonnet 4.6, GPT-5.2, and Claude Haiku 4.5) achieves 124 percent higher merged-archive QD-Score (45.90 vs. 20.46) and 28 percent higher coverage (80.6 percent vs. 63.0 percent of cells) than a single-node baseline at equal total LLM-call budget. The heterogeneous ensemble also outperforms an equally-budgeted homogeneous ensemble on QD-Score, coverage, and held-out solution generality across all four model families. These results provide the first empirical evidence that model diversity, not merely parallelism, is the key driver of gain in distributed LLM-based QD search.</p> + https://arxiv.org/abs/2605.27130 Tue, 26 May 2026 00:00:00 +0000 - Augmenting Attention with Exponentially Decaying Memory Improves Query-Aware KV Sparsity - https://arxiv.org/abs/2605.28640 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.28640.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiuying Wei, Caglar Gulcehre</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Efficient inference is critical for long-context language models, where attention computation and KV-cache access dominate the cost. Recent work RAT+, introduces a recurrence-augmented attention backbone that enables flexible dilated attention at inference time. In this paper, we investigate whether this exponentially decaying memory can also improve existing query-aware sparse inference methods. Using representative methods including Quest, MoBA, and SnapKV, we show that RAT+ consistently improves accuracy over standard attention across sparse budgets on eight needle-in-a-haystack tasks. We validate these gains both on the released checkpoints from the RAT+ paper and on OLMo2-7B, which we continue pretraining with the added memory module for 10B tokens. Finally, we propose two hypotheses explaining why this memory module benefits query-aware sparse inference and design targeted experiments to support them.</p> - https://arxiv.org/abs/2605.28640 + Pruning and Distilling Mixture-of-Experts into Dense Language Models + https://arxiv.org/abs/2605.28207 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.28207.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Junhyuck Kim, Jihun Yun, Haechan Kim, Gyeongman Kim, Joonghyun Bae, Jaewoong Cho</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Mixture-of-Experts (MoE) is now the dominant architecture for frontier language models, yet it requires all expert parameters to be loaded in memory, making it less preferable for memory-constrained deployment. Existing compression methods reduce the number of experts but the output remains an MoE model with the same fundamental limitation. We present the first systematic framework for converting a trained MoE into a standard fully dense architecture: experts are scored, selected, and grouped, then concatenated into a dense FFN and refined by knowledge distillation from the MoE teacher. We evaluate 7 scoring, 5 grouping, and 2 magnitude scaling methods across a range of selected expert counts on Qwen3-30B-A3B, yielding 350 configurations. We find that the choice of scoring method is the most impactful, with our novel diversity-aware scoring consistently outperforming prior methods on Qwen3-30B-A3B, DeepSeek-V2-Lite, and GPT-OSS-20B. Under a controlled comparison at matched parameter count, MoE-to-dense outperforms dense-to-dense pruning by +6.3 pp in average downstream accuracy after ~4B-token distillation at 1.6x faster training wall-clock speed.</p> + https://arxiv.org/abs/2605.28207 Wed, 27 May 2026 00:00:00 +0000 - CORE: Contrastive Reflection Enables Rapid Improvements in Reasoning - https://arxiv.org/abs/2605.28742 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.28742.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Linas Nasvytis, Simon Jerome Han, Ben Prystawski, Satchel Grant, Noah D. Goodman, Judith E. Fan</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Language models can use verifiable rewards to improve at a wide variety of reasoning tasks. However, both parametric (e.g. RLVR) and non-parametric (e.g. prompt optimization) approaches to doing so typically require hundreds of training samples and thousands of model rollouts, making them expensive in the best case and intractable in the worst. To address this challenge, we introduce Contrastive Reflection (CORE), a non-parametric learning algorithm that compares past reasoning traces to generate insights: short natural-language descriptions of reasoning strategies and constraints that capture differences between successful and unsuccessful problem attempts. Across four reasoning tasks, we demonstrate that CORE enables more rapid improvement than both parametric (GRPO) and non-parametric (GEPA, episodic RAG, and MemRL) methods, while using fewer rollouts. Under fixed rollout budgets with as few as five training samples, we then show that CORE also achieves comparable or greater performance gains than each baseline. Finally, we highlight how CORE is also substantially more context-efficient than non-parametric baselines, requiring fewer prompt tokens while storing learned knowledge as compact, interpretable natural-language insights. Our results therefore suggest that distilling contrasts between successful and unsuccessful reasoning traces into abstract and useful insights can provide a more efficient and interpretable route to model self-improvement than weight updates, prompt optimization, or direct reuse of stored reasoning traces.</p> - https://arxiv.org/abs/2605.28742 - Wed, 27 May 2026 00:00:00 +0000 - - - Towards Human-Like Interactive Speech Recognition With Agentic Correction and Semantic Evaluation - https://arxiv.org/abs/2605.29430 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.29430.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zixuan Jiang, Yanqiao Zhu, Peng Wang, Qinyuan Chen, Xinjian Zhao, Xipeng Qiu, Wupeng Wang, Zhifu Gao, Xiangang Li, Kai Yu, Xie Chen</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Automatic speech recognition (ASR) is a core component of human--computer interaction and an increasingly important front-end for LLM-based assistants and agents. However, most current ASR systems still follow a single-pass paradigm, which is poorly aligned with human communication, where misunderstandings are resolved through iterative clarification and refinement. This mismatch makes it difficult to correct meaning-critical errors once they occur. Meanwhile, token-level metrics such as WER or CER cannot adequately reflect such a problem. To address these limitations, we formulate Interactive ASR as a multi-turn refinement task and propose Agentic ASR, a closed-loop framework that combines a single-pass ASR front-end with semantic correction, intent routing, and reasoning-based editing. We further introduce the Sentence-level Semantic Error Rate (S^2ER), an LLM-based semantic evaluation metric, together with an Interactive Simulation System for scalable and reproducible benchmarking. Experiments on multilingual, named-entity-intensive, and code-switching benchmarks show that iterative interaction consistently reduces semantic errors, with much larger gains in S^2ER than in conventional token-level metrics. Human--AI alignment and ablation studies further validate the reliability of the semantic judge and the robustness of the proposed framework. The code is available at: https://interactiveasr.github.io/ and the live demo is available at https://i-asr.sjtuxlance.com/</p> - https://arxiv.org/abs/2605.29430 - Thu, 28 May 2026 00:00:00 +0000 - - - The Distillation Game: Adaptive Attacks & Efficient Defenses - https://arxiv.org/abs/2605.22737 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.22737.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Youssef Allouah, Mahdi Haghifam, Sanmi Koyejo, Reza Shokri</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Distillation attacks create a deployment trade-off for model providers: the same outputs that make a model more useful can also make it easier to imitate. We study this trade-off through a minimax game between a utility-constrained teacher and an adaptive student. Our framework yields tractable one-sided response rules: an adaptive evaluation rule in which the student reweights high-value examples, and a teacher-side defense template that suppresses outputs most useful for distillation. From a cheap proxy for example value, we derive Product-of-Experts (PoE), a simple forward-pass-only defense that combines the teacher with a proxy student during generation. Empirically, adaptive evaluation reveals a large passive--adaptive gap: on state-of-the-art defenses, adaptive students recover substantially more capability than passive evaluation suggests on GSM8K and MATH. Under this stronger evaluation, the apparent robustness gap between expensive defenses and PoE narrows considerably, while PoE remains substantially cheaper and preserves higher-quality reasoning traces. Overall, our results suggest that strong distillation remains difficult to stop, and that progress on antidistillation should be judged against adaptive students rather than passive ones. Our code is available at: https://github.com/ysfalh/distillation-game.</p> - https://arxiv.org/abs/2605.22737 + Send a SCOUT First: Pre-hoc Reasoning for Adaptive Detector Allocation in Prompt-Injection Defense + https://arxiv.org/abs/2605.30837 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.30837.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuhao Zhang, Jiarui Li, Qi Cao, Ruiyi Zhang, Pengtao Xie</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Prompt-injection detectors are heterogeneous: each is strong on a different slice of attacks, and none is always reliable. Yet existing systems still treat detection as a fixed single-detector pipeline, committing every request to one detector's blind spots. We reframe defense as detector allocation: given a heterogeneous pool, decide per request which detectors to run and whether to escalate to an LLM judge. Our framework SCOUT (Scalable and Controllable Outcome-prediction for Uncertainty-aware Triage) makes this decision dynamic by predicting each detector's per-sample reliability and latency from how it behaved on similar past inputs, and exposes a single safety-utility threshold to the operator (where utility bundles benign-pass rate and wall-clock). To evaluate this setting, we build SCOUT-450, a benchmark that captures the structurally complex, agent-facing injections that older prompt-injection sets under-represent. On SCOUT-450, a safety-oriented operating point reduces attack-success rate by 46% and total wall-clock by 40% relative to an always-on GPT-4o judge, at a 5.1-point benign-utility drop. SCOUT also transfers to three external benchmarks (BIPIA, IPI, and IHEval), improving the safety-utility frontier.</p> + https://arxiv.org/abs/2605.30837 Fri, 29 May 2026 00:00:00 +0000 - Critic-R: Improving Agentic Search using Instruction-tuned Retrievers with Natural Language Introspective Feedback - https://arxiv.org/abs/2606.00590 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.00590.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Md Zarif Ul Alam, Alireza Salemi, Hamed Zamani</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Agentic search systems iteratively interact with retrieval models to answer complex queries. Despite substantial progress, optimizing retrievers for agentic search remains challenging, often requiring heavy co-training or gold-standard annotations that limit real-world applicability. We propose Critic-R, a framework that explicitly closes the feedback loop between the reasoning agent and the retrieval model during both inference and training. Critic-R introduces a critic model that evaluates the agent's introspective reasoning trace after consuming retrieved evidence to determine whether the retrieved context sufficiently supports the next reasoning step. Critic-R has two complementary mechanisms: Critic-R-Zero, an inference-time query refinement loop that iteratively rewrites queries and retrieval instructions, and Critic-Embed, an optimization approach for retrieval models that leverages successful and failed refinement trajectories as automatic supervision without requiring manual relevance annotation. We evaluate Critic-R on HotpotQA, 2WikiMultihopQA, MuSiQue, and Bamboogle. Results show that Critic-R significantly improves both retrieval quality and downstream answer accuracy.</p> - https://arxiv.org/abs/2606.00590 + Human Psychometric Questionnaires Mischaracterize LLM Behavior + https://arxiv.org/abs/2509.10078 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2509.10078.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Woojung Song, Dongmin Choi, Yoonah Park, Jongwook Han, Eun-Ju Lee, Yohan Jo</p><p><b>Upvotes:</b> 31</p><p><b>Summary:</b> We examine whether human psychometric questionnaires can serve as reliable tools for characterizing and predicting LLM behavior in everyday user interactions. We analyze eight open-source LLMs by comparing their value and personality profiles derived from two different methods: Likert self-reports on established questionnaires (PVQ-40/21 and BFI-44/10) and generation probabilities over value-laden responses to everyday user queries. The two profiles diverge substantially. Within-construct item consistency, often cited as evidence of stable LLM dispositions, disappears in generation probabilities. We attribute this gap to the fact that explicit lexical cues in established questionnaire items allow models to recognize the target construct and respond in alignment-consistent, socially desirable ways, whereas realistic user queries provide no such cues. In addition, demographic persona prompts shift models' responses to human questionnaires in ways consistent with real human patterns, but no such shifts appear in the generation probabilities of responses to realistic user queries, showing their limited ability to simulate the behaviors of target demographics in real-world user interactions. Overall, our study shows that human psychometric questionnaires are insufficient tools for predicting LLM behavior and suggests generation-based profiling as a more accurate measure.</p> + https://arxiv.org/abs/2509.10078 + Fri, 29 May 2026 00:00:00 +0000 + + + SDR: Set-Distance Rewards for Radiology Report Generation + https://arxiv.org/abs/2606.00440 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.00440.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Halil Ibrahim Gulluk, Max Van Puyvelde, Wim Van Criekinge, Olivier Gevaert</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Reinforcement learning with verifiable rewards has rapidly advanced reasoning in vision--language models. However, for chest X-ray report generation, the standard rewards (i.e. exact-match accuracy and step-level processes) are incompatible because the reports consist of unordered and orthogonal findings, rather than a causal reasoning chain. We address this gap with a set-based view: each report is split into sentences and embedded by a frozen sentence transformer, yielding unordered embedding sets. We propose the use of set-to-set distances between generated and reference embeddings as continuous, permutation-invariant rewards. Across two datasets and three vision--language models (Qwen3-VL-2B/4B, Gemma3-4B), post-training with set-to-set distance based rewards via GRPO consistently outperforms supervised fine-tuning and exact-match GRPO on all headline metrics (BERTScore, RadGraph F1 and CheXbert F1 by average \%6.80, \%7.82 and \%4.45 relative improvements respectively). The same set distances also enable test-time best-of-N selection: scoring candidates by their distance to training-report embeddings outperforms random selection on our trained models as well as three closed-source LLMs (Mistral-Small, Gemini-2.5 Flash-Lite, GPT-4o-mini) with on average \%16.4 relative improvement on BERTScore. Used as a streaming signal, they support a more efficient form of test-time scaling: pruning low-scoring candidates mid-generation reduces generated tokens by over 50\% while preserving the Findings quality of full best-of-N selection. Together these results establish set-distance rewards as a unified signal for both post-training and test-time scaling in chest X-ray report generation. Our code is publicly https://anonymous.4open.science/r/Set-Distance-Rewards-CXR-BFDA{available}.</p> + https://arxiv.org/abs/2606.00440 Sat, 30 May 2026 00:00:00 +0000 - Parametric Social Identity Injection and Diversification in Public Opinion Simulation - https://arxiv.org/abs/2603.16142 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2603.16142.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hexi Wang, Yujia Zhou, Bangde Du, Qingyao Ai, Yiqun Liu</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Large language models (LLMs) have recently been adopted as synthetic agents for public opinion simulation, offering a promising alternative to costly and slow human surveys. Despite their scalability, current LLM-based simulation methods fail to capture social diversity, producing flattened inter-group differences and overly homogeneous responses across demographic groups. We identify this limitation as a Diversity Collapse phenomenon in LLM hidden representations, where distinct social identities become increasingly indistinguishable across layers. Motivated by this observation, we propose Parametric Social Identity Injection (PSII), a general framework that injects explicit, parametric representations of demographic attributes and value orientations directly into intermediate hidden states of LLMs. Unlike prompt-based persona conditioning, PSII enables fine-grained and controllable identity modulation at the representation level. Extensive experiments on the World Values Survey using multiple open-source LLMs show that PSII significantly improves distributional fidelity and diversity, reducing KL divergence to real-world survey data while enhancing overall diversity. This work provides new insights into representation-level control of LLM agents and advances scalable, diversity-aware public opinion simulation.</p> - https://arxiv.org/abs/2603.16142 - Mon, 01 Jun 2026 00:00:00 +0000 + Honest Lying: Understanding Memory Confabulation in Reflexive Agents + https://arxiv.org/abs/2605.29463 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2605.29463.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Prakhar Dixit, Sadia Kamal, Tim Oates</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Reflexion-style agents rely on self-generated reflections as memory, implicitly assuming that agents can accurately diagnose their own failures. We show that this assumption can fail systematically: across ALFWorld and HumanEval, agents store confident but incorrect interpretations of the task and continue acting on them across trials, even though the environment resets to the correct task each time. We call this failure mode memory confabulation and introduce the Reflection Repetition Rate (RRR), a log-based metric that detects repeated reliance on incorrect reflective content. Using RRR, we identify 16 frozen environments in ALFWorld, where 0 of 121 reflections mention the correct target object, and 4 analogous cases in HumanEval. Our mitigation replaces open-ended self-diagnosis with programmatic extraction of trajectory-level failure signals, increasing correct object mention from 0% to 86%, reducing RRR from 0.64 to 0.10, and solving 3 of 16 frozen ALFWorld environments, suggesting that reflective memory can reinforce false beliefs rather than correct them.</p> + https://arxiv.org/abs/2605.29463 + Sun, 31 May 2026 00:00:00 +0000 - LayerRoute: Input-Conditioned Adaptive Layer Skipping via LoRA Fine-Tuning for Agentic Language Models - https://arxiv.org/abs/2606.01838 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.01838.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Prateek Kumar Sikdar</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Agentic language model systems alternate between two structurally distinct step types: structured tool calls (short, deterministic, low perplexity) and open-ended planning/reasoning steps (long, complex, high perplexity). Despite this heterogeneity, current inference systems apply identical compute to every step. We introduce LayerRoute, a lightweight adapter that learns to selectively skip transformer blocks on a per-input basis. LayerRoute augments each of the 24 transformer blocks in Qwen2.5-0.5B-Instruct with: (1) a per-layer router (~897 parameters, Linear(896,1)) that outputs a hard binary gate via the straight-through estimator, and (2) LoRA adapters (rank 8, ~1.08M parameters) on the Q/K/V/O attention projections. The backbone weights remain frozen. A single end-to-end training pass on agentic data (Hermes, Glaive, GSM8K, Turing) with a gate regularisation term forces the system to discover which blocks are skippable per input type. After 3,000 steps (6.4 minutes on an A100 40GB), LayerRoute achieves a 12.91% skip differential: tool calls skip 15.25% of FLOPs while planning steps skip only 2.34%, using only 1.10M trainable parameters (0.22% of the 494M backbone). Quality improves over the base model due to LoRA adaptation, with perplexity delta of -1.29 on tool calls and -1.30 on planning.</p> - https://arxiv.org/abs/2606.01838 - Mon, 01 Jun 2026 00:00:00 +0000 + Trust Functions: Near-Lossless Weak-to-Strong Generalization by Learning When to Trust the Weak Teacher + https://arxiv.org/abs/2606.01000 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.01000.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Arda Uzunoglu, Alvin Zhang, Daniel Khashabi</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Weak-to-strong generalization studies how to improve a strong student using supervision from a weaker teacher when reliable labels are scarce. We view this primarily as a data selection problem, where the key challenge is to identify which weak labels are reliable enough to serve as a training signal. To address this, we introduce trust functions that assign each weak label a scalar trust score and use these scores to filter weak supervision. Across several domains, including world knowledge, quantitative reasoning, and strategy games, trust filtering yields students that match and sometimes surpass ground-truth supervision, achieving near-lossless weak-to-strong generalization. Moreover, trust functions enable an iterative weak-to-strong chain that compounds gains by training a student and reusing it as the next teacher, amplifying the gains. There are several mechanisms to which advantage of trust functions can be attributed.</p> + https://arxiv.org/abs/2606.01000 + Sun, 31 May 2026 00:00:00 +0000 - HarnessForge: Joint Harness and Policy Evolution for Adaptive Agent Systems - https://arxiv.org/abs/2606.01779 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.01779.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingju Chen, Can Lv, Guibin Zhang, Heng Chang, Shiji Zhou</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> LLM agents are increasingly expected to operate across heterogeneous task regimes that require distinct execution paradigms. This challenges fixed agent systems and motivates system-level meta-adaptation beyond isolated component updates. While existing works have adapted external harness or trained underlying reasoning policies, full-system adaptation remains insufficiently characterized. The adaptation space between structure and execution is rarely made explicit, and the compatibility between the external harness and the internal reasoner is not optimized jointly. We propose HarnessForge, a meta-adaptive framework for evolving LLM agent systems. HarnessForge formulates an agent system as a harness--policy pair, defining a stable adaptation space that separates harness-level execution structure from policy-level reasoning behavior. It then performs harness--policy co-evolution through fault-guided harness tailoring and harness-conditioned policy alignment. Experiments across five benchmarks from diverse domains show that HarnessForge consistently improves both Qwen3-4B and Qwen3-8B backbones, outperforming harness-only and policy-only baselines with gains of up to 12.0\% over the strongest baseline and achieving favorable rollout-efficiency tradeoffs, demonstrating that harness--policy co-evolution is effective, and that executable compatibility between the harness and reasoning policy is essential for agent-system adaptation. The code is available at https://github.com/mingju-c/HarnessForge.</p> - https://arxiv.org/abs/2606.01779 - Mon, 01 Jun 2026 07:00:35 +0000 - - - A Cookbook of 3D Vision: Data, Learning Paradigms, and Application - https://arxiv.org/abs/2606.04291 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.04291.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hongyang Du, Zongxia Li, Dawei Liu, Runhao Li, Haoyuan Song, Qingyu Zhang, Yubo Wang, Jingcheng Ni, Shihang Gui, Congchao Dong, Tao Hu</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> 3D vision has rapidly evolved, driven by increasingly diverse data representations, learning paradigms, and modeling strategies. Yet the field remains fragmented across representations and benchmarks, making it difficult to develop unified perspectives on efficiency, fidelity, and scalability. This work provides a data-centric taxonomy of 3D vision that connects geometric representations, datasets, learning frameworks, and applications within a single conceptual map. We begin by analysing the principal structural representations of 3D data--point clouds, meshes, voxels, and 3D Gaussians--along with their acquisition pipelines. We then examine how dataset design, benchmark construction, and supervision regimes shape recent advances, spanning 2D-supervised 3D learning, implicit neural representations, and 4D world modeling. Through this integrative lens, we clarify the relationships among representations, learning paradigms, and downstream tasks in reconstruction, generation, and video modeling, offering a consolidated view of emerging trends toward balancing efficiency and fidelity and toward multimodal geometric grounding.</p> - https://arxiv.org/abs/2606.04291 + Reasoning over Grammar: Can Synthetic Linguistic Reasoning Traces Enhance Low-Resource Machine Translation? + https://arxiv.org/abs/2606.03782 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.03782.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Renhao Pei, Yihong Liu, Sampo Pyysalo, Hinrich Schütze, Shaoxiong Ji</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Large language models (LLMs) offer a promising approach to machine translation (MT) for extremely low-resource languages by incorporating linguistic resources through in-context learning. However, LLMs often struggle to apply grammatical information effectively during translation. Inspired by recent progress in chain-of-thought reasoning, we investigate whether low-resource MT can benefit from structured intermediate steps of linguistic analysis and grammatical reasoning. We propose a pipeline for automatically generating step-by-step linguistic reasoning traces from Universal Dependencies treebanks, dictionaries, and grammar-rule banks. We evaluate these traces in three settings: in-context learning (ICL), supervised fine-tuning (SFT), and reinforcement fine-tuning (RFT), on Xibe and Chintang as test cases. Our results show that linguistic reasoning traces are most effective as inference-time guidance: in ICL, reliable sentence-specific traces substantially improve translation performance across most models, languages, and metrics. In contrast, using the linguistic reasoning traces as training data yields smaller and less consistent gains, as models learn the trace format but often generate erroneous content. These findings suggest that LLMs can leverage grammatical information for low-resource MT when given reliable linguistic analyses, while learning to generate such analyses remains a major bottleneck.</p> + https://arxiv.org/abs/2606.03782 Tue, 02 Jun 2026 00:00:00 +0000 - Measuring Model Robustness via Fisher Information: Spectral Bounds, Theoretical Guarantees, and Practical Algorithms - https://arxiv.org/abs/2606.04767 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.04767.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chong Zhang, Xiang Li, Jia Wang, Qiufeng Wang, Xiaobo Jin</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> The robustness of deep neural networks is crucial for safety-critical deployments, yet existing evaluation methods are often attack-dependent and lack interpretability. We propose a principled, attack-agnostic robustness metric based on the spectral norm of the Fisher Information Matrix (FIM), which quantifies the worst-case sensitivity of the model's output distribution to input perturbations. Theoretically, we establish that the FIM equals the variance of the input Jacobian and derive closed-form spectral bounds for common architectures, including VGG, ResNet, DenseNet, and Transformer, providing the first theoretical robustness ranking. To enable scalable evaluation, we develop efficient algorithms, including power iteration and Hutchinson-based estimation, that support both white-box and black-box settings. Extensive experiments across multiple datasets, including CIFAR, ImageNet, and medical images, and across multiple architectures show a strong correlation between our metric and adversarial vulnerability. Our framework serves as an interpretable diagnostic tool that complements attack-based evaluations, offering insights into architectural sensitivity and guiding the design of more robust models. Code is available at: https://github.com/franz-chang/SRP/.</p> - https://arxiv.org/abs/2606.04767 + Lean4Agent: Formal Modeling and Verification for Agent Workflow and Trajectory + https://arxiv.org/abs/2606.06523 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06523.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ruida Wang, Jerry Huang, Pengcheng Wang, Xuanqing Liu, Luyang Kong, Tong Zhang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Equipping Large Language Models (LLMs) to execute reliable multi-step workflows has become a central challenge in artificial intelligence. Despite recent advances in LLMs' agentic capabilities, most agent systems still lack formal methods for specifying, verifying, and debugging their workflow and execution trajectories. This challenge mirrors a long-standing problem in mathematics, where the ambiguity of natural languages (NLs) motivates the development of formal languages (FLs). Inspired by this paradigm, we propose **Lean4Agent**, to the best of our knowledge, the first framework that uses Lean4, a dependent-type FL to model and verify agent behavior. **Lean4Agent** launches **FormalAgentLib**, an extensible Lean4 library for formally modeling and verifying agent workflows' semantic consistency under explicit assumptions, and enabling localization of execution-time failures revealed by trajectories. Building on **FormalAgentLib**, we further develop **LeanEvolve**, which applies results in **FormalAgentLib** to revise workflows to enhance its capability. Extensive experiments on a hard problem subset of SWE-Bench-Verified and a subset of ELAIP-Bench across 5 leading LLMs indicate that the verification-passing workflows outperform the failing ones by an average of **11.94%**, and **LeanEvolve** further improves SWE performance by **7.47%** on average. Furthermore, **Lean4Agent** establishes a foundation for a new field of using expressive dependent-type FL to formally model and verify agent behavior.</p> + https://arxiv.org/abs/2606.06523 + Tue, 02 Jun 2026 00:00:00 +0000 + + + Text-to-Image Models Need Less from Text Encoders Than You Think + https://arxiv.org/abs/2606.03715 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.03715.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Nurit Spingarn, Noa Cohen, Tamar Rott Shaham, Tomer Michaeli</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Text-to-image models rely on text prompts as their primary interface to human intent. Prompts are encoded by a text encoder into embeddings that condition the image generation process. Beyond individual token meanings, text embeddings encode contextual information across the full prompt, such as compositionality and attribute binding. However, whether image models actually exploit this richer information remains underexplored. Here, we address the question: Which aspects of text representation are essential for image generation? We show that text-to-image diffusion transformer-based models commonly rely only on two relatively straightforward aspects of text representations: (i) the merging of adjacent tokens into a word representation, for words spanning multiple tokens, and (ii) word order, which is imprinted by the positional embedding of the text-encoder. To show this, we construct a new text embedding that encodes only individual word meanings and order but lacks any contextual information about the full prompt. We find that this bag of position-tagged words representation is sufficient to successfully guide image generation, achieving visual quality and text fidelity that are on par with full text embedding-guided generation. This demonstrates that, contrary to common belief, text-to-image models often do not use the rich information encoded in the text embedding beyond individual word meanings and word order. Instead, the decoding of complex linguistic structures is performed by the image model itself. Project webpage: https://nsping13.github.io/contextless-TTI/</p> + https://arxiv.org/abs/2606.03715 + Tue, 02 Jun 2026 00:00:00 +0000 + + + Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill + https://arxiv.org/abs/2606.03980 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.03980.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Tao Chen, Gangwei Jiang, Pengyu Cheng, Siyuan Huang, Yihao Liu, Jingwei Ni, Jiaqi Guo, Mengyu Zhou, Kai Tang, Junling Liu, Qinliang Su, Xiaoxi Jiang, Guanjun Jiang</p><p><b>Upvotes:</b> 10</p><p><b>Summary:</b> Reward models (RMs) provide critical feedback signals for LLM post-training, notably in reinforced fine-tuning (RFT) and reinforcement learning (RL) pipelines. However, current reward evaluation relies on heterogeneous criteria such as rule-based verifiers, ground-truth references, procedural checklists, and complex rubrics, where a unified mechanism to integrate all types of evidence remains unexplored. To this end, we propose Skill Reward Model (Skill-RM), a unified framework that reformulates reward modeling as the execution of a reusable Reward-Evaluation Skill. By treating reward computation as a structured agentic task, Skill-RM provides a consistent interface to orchestrate heterogeneous resources, dynamically selecting and aggregating evidence tailored to the specific requirements of each input. This approach enables the reward model to move beyond static evaluation, ensuring consistency and transparency across diverse tasks. Extensive experiments on reward benchmarks and downstream applications, including best-of-N selection and reinforcement learning, demonstrate that Skill-RM consistently outperforms traditional judge baselines. Our findings suggest that Skill-RM not only provides a unified solution for reward modeling but also achieves superior performance through the strategic and dynamic orchestration of evidence. The code is at https://github.com/Qwen-Applications/Skill-RM.</p> + https://arxiv.org/abs/2606.03980 + Tue, 02 Jun 2026 00:00:00 +0000 + + + CIPER: A Unified Framework for Cross-view Image-retrieval and Pose-estimation + https://arxiv.org/abs/2606.05011 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05011.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yurim Jeon, Dongseong Seo, Seung-Woo Seo</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Cross-view geo-localization estimates the geographic location of a ground image by matching it against an aerial image database. Existing methods tackle this through either large-scale retrieval or precise pose estimation, but not both: retrieval-based methods enable wide-area search at the cost of localization accuracy, while pose estimation methods achieve high precision within only a narrow search space. Naively cascading these pipelines introduces error propagation and inconsistent feature representations. We formulate cross-view geo-localization as a unified problem requiring simultaneous city-scale retrieval and precise 3-DoF pose estimation. We propose CIPER (Cross-view Image-retrieval and Pose-estimation transformER), a single architecture that jointly performs both tasks through mutually beneficial feature learning. CIPER uses a shared transformer encoder with task-specific tokens to disentangle global retrieval features from spatial localization cues. To bridge the large domain gap between ground and aerial views, we introduce a two-way transformer pose decoder that uses ground features as spatial queries for bidirectional cross-attention. A set prediction strategy further enables stable 3-DoF regression under a unified multi-task objective. Experiments on VIGOR, KITTI, and Ford Multi-AV demonstrate competitive performance, especially under limited field-of-view and arbitrary orientation conditions. Code is available at https://github.com/yurimjeon1892/CIPER.</p> + https://arxiv.org/abs/2606.05011 Wed, 03 Jun 2026 00:00:00 +0000 - Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models - https://arxiv.org/abs/2606.03988 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.03988.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mahtab Bigverdi, Linjie Li, Weikai Huang, Yiming Liu, Jaemin Cho, Jieyu Zhang, Tuhin Kundu, Chris Dangjoo Kim, Zelun Luo, Linda Shapiro, Ranjay Krishna</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Vision language models (VLMs) excel at many tasks but still struggle with spatial reasoning when critical information is not directly observable. Many such problems require imaginative perception: inferring what would be seen from an unseen viewpoint, tracing paths through occluded spaces, or integrating partial observations into a coherent spatial representation. We introduce Imaginative Perception Tokens (IPT), intermediate perceptual representations that externalize what a VLM would perceive under alternative spatial configurations while remaining consistent with the observed input. To study this capability, we formulate three tasks, Perspective Taking (PET), Path Tracing (PT), and Multiview Counting (MVC), and construct datasets of approximately 20K examples with ground truth imaginations, answers, and evaluation benchmarks. Using the unified VLM BAGEL as the backbone, IPT supervision consistently improves spatial reasoning and often outperforms textual chain of thought training, even without generating images at inference time. On MVC, IPT improves accuracy by 3.4% and achieves competitive performance with strong closed-source models on PT. We further find that combining IPT and label-only supervision yields additional gains, whereas textual chain of thought can substantially degrade performance, suggesting a modality mismatch when spatial computation is forced through language. Overall, IPT provides a principled supervision signal for reasoning about unobserved spatial structure, improving generalization while producing interpretable intermediate representations.</p> - https://arxiv.org/abs/2606.03988 + Self-Evaluation Is Already There: Eliciting Latent Judge Calibration in Base LLMs with Minimal Data + https://arxiv.org/abs/2606.05122 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05122.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> XiuYu Zhang, Yi Shan, Junfeng Fang, Zhenkai Liang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Large language models are increasingly evaluated by other models, raising a natural question: can a model predict how a judge will score its own output? We find that the ability is largely present before any targeted training: prompted few-shot, a base model already predicts an external judge's multi-attribute quality scores on open-ended responses well above chance across three benchmarks. We introduce Self-Evaluation Elicitation (SEE), a method that surfaces this latent ability through a short cycle comprising a calibration-coupled reinforcement learning phase that improves the answer and predicts the judge, followed by a masked distillation phase that sharpens the prediction while leaving the answer untouched. From 160 unique examples, roughly 31x fewer than a reinforcement learning baseline, SEE improves held-out calibration across three benchmarks while preserving answer quality. The elicited self-evaluation is sharply localized within the model's own token distribution and stable across judges it was never trained against, indicating a transferable notion of quality rather than a single judge's preference. These results reframe judge-aligned self-evaluation as a problem of elicitation rather than acquisition.</p> + https://arxiv.org/abs/2606.05122 Wed, 03 Jun 2026 00:00:00 +0000 - Reinforcement Learning from Rich Feedback with Distributional DAgger - https://arxiv.org/abs/2606.05152 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05152.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Rishabh Agrawal, Jacob Fein-Ashley, Paria Rashidinejad</p><p><b>Upvotes:</b> 3</p><p><b>Summary:</b> Reasoning models have advanced rapidly, but the dominant reinforcement learning from verifiable rewards (RLVR) recipe remains surprisingly narrow: sample many responses and reward each with a single bit indicating whether the final answer is correct. Yet many settings provide rich feedback, including execution traces, tool outputs, expert corrections, and model self-evaluations. We study how to use such feedback through a distributional variant of the classic imitation learning algorithm DAgger, where the learner has local access to an expert distribution on states visited by the current policy. This yields a simple forward cross-entropy objective that admits a blackbox expert and whose sequence-level gradient {conduct rich credit assignment by propagating} future expert-student disagreement back to earlier decisions. We show that prior RL with self-distillation objectives based on reverse KL or Jensen-Shannon fail to guarantee monotonic policy improvement: even when the expert has higher reward, their updates may increase probability on worse actions. In contrast, we show that forward cross-entropy admits monotonic policy improvement and enjoys guarantees on regret. We further show that our objective optimizes a lower bound on teacher-weighted likelihood of success, leading to improved Pass@N. Empirically, our approach, DistIL, improves over RLVR and RL with self-distillation baselines across a variety of domains: scientific reasoning, coding, and solving hard mathematical problems.</p> - https://arxiv.org/abs/2606.05152 + Why Muon Outperforms Adam: A Curvature Perspective + https://arxiv.org/abs/2606.04662 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.04662.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shuche Wang, Fengzhuo Zhang, Jiaxiang Li, Dirk Bergemann, Zhuoran Yang</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Muon improves training efficiency over Adam in large language-model training by about two times, but the local geometric source of this advantage remains unclear. Our work takes a first step toward demystifying Muon's superiority over Adam from a curvature perspective. First, we apply a second-order Taylor approximation to the training landscape and show that Muon achieves a larger one-step loss decrease than Adam at matched validation loss. The two optimizers have comparable first-order gains, but Muon consistently incurs a smaller second-order curvature penalty. Second, we decompose this curvature penalty into the squared update norm and Normalized Directional Sharpness (NDS). We find that Muon and Adam have comparable update norms, so Muon's smaller curvature penalty is driven by lower NDS, not update scale. Third, we study how training data and model structure shape Muon's NDS advantage. Using Zipf-Probabilistic Context-Free Grammar (PCFG) data with controlled imbalance, we show that data imbalance amplifies Muon's NDS advantage over Adam. A within-/cross-layer decomposition further shows that, in the middle and late stages of training, Muon's lower NDS is mainly sustained by smaller within-layer curvature. Beyond empirical evidence, we analyze stylized quadratic problems with heterogeneous curvature and gradient alignment toward high-curvature modes. We prove that Muon attains a smaller average NDS than GD by balancing update energy across curvature groups; when curvature heterogeneity is sufficiently strong, this also yields lower local quadratic loss after the same number of steps.</p> + https://arxiv.org/abs/2606.04662 Wed, 03 Jun 2026 00:00:00 +0000 - GENEB: Why Genomic Models Are Hard to Compare - https://arxiv.org/abs/2606.04525 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.04525.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Daria Ledneva, Mikhail Nuridinov, Denis Kuznetsov</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> Progress in genomic foundation models is difficult to assess due to fragmented benchmarks, incompatible evaluation protocols, and task-specific reporting. As a result, claims of superiority or generality across models are often not directly comparable. We introduce GENEB, a large-scale diagnostic benchmark that evaluates frozen representations from 40 genomic foundation models across 100 tasks spanning 13 functional categories under a unified probing-based protocol, including few-shot regimes. GENEB enables controlled comparison across model scale, architecture, tokenization, and pretraining data while explicitly exposing task-level trade-offs. Our analysis shows that aggregate leaderboards are unstable: model rankings vary sharply across task categories, scale provides only modest and inconsistent gains, and architectural and pretraining alignment frequently outweigh parameter count. These results highlight limitations of current evaluation practices and position GENEB as a reference framework for principled comparison and category-aware model selection in genomic machine learning.</p> - https://arxiv.org/abs/2606.04525 + Agents' Last Exam + https://arxiv.org/abs/2606.05405 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05405.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yiyou Sun, Xinyang Han, Weichen Zhang, Yuanbo Pang, Tianyu Wang, Yuhan Cao, Yixiao Huang, Chris Duroiu, Haoyun Zhang, Jeffrey Lin, Weishu Zhang, Tyler Zeng, Ying Yan, Bo Liu, Hanson Wen, Mingyang Xu, Xiaoyuan Liu, Zimeng Chen, Weiyan Shi, Amanda Dsouza, Vincent Sunn Chen, Patrick Bryant, Carl Boettiger, Yamini Rangan, Bradley Rothenberg, Kyle Steinfeld, Arvind Rao, Tapio Schneider, Georgios Yannakakis, Laure Zanna, Kaan Ozbay, Ida Sim, Tarek Zohdi, George Em Karniadakis, Jack Gallant, Teresa Head-gordon, Yushan Li, Wenxi Deng, Tao Sun, Huiqi Wang, Zhun Wang, Justin Xu, Chris Yuhao Liu, Yafei Cheng, Rongwang Hu, Aras Bacho, Shengcao Cao, Zengyi Qin, Yixiong Chen, Hengduan Fan, Hao Liu, Lin Zeng, Shashank Muralidhar Bharadwaj, Litian Gong, Yingxuan Yang, Maojia Song, Ruheng Wang, Zongzheng Zhang, Honglin Bao, Shuo Lu, Jianhong Tu, Zhonghua Wang, Zheng Zhang, Zijiao Chen, yanqiong Jiang, Zhendong Li, Bohan Lyu, Chang Ma, Peiran Xu, Benran Zhang, Shangding Gu, Haoyue Hua, Haoyang Li, Wanzhe Liao, Chengzhi Liu, Junbo Peng, Haoran Sun, Zechen Xu, Bo Chen, Jiayi Cheng, Yi Jiang, Keying Kuang, Yuan Li, Youbang Pan, Ziyan Rao, Alexander Schubert, Yifan Shen, Vincent Siu, Xiatao Sun, Kangqi Zhang, Xiaopan Zhang, Yuchen Zhu, Ishaan Singh Chandok, Lei Ding, Jingxuan Fan, Andrew Glover, Jiaming Hu, Yiran Hu, Wenbo Huang, Zixin Jiang, Haoran Jin, Lukas Kim, Ming Liu, Yang Liu, Alireza Rafiei, Xuhuan Shen, Kunyang Sun, Sophia Sun, Ting Sun, Eric Wang, Yixin Wang, Hanwen Xing, Sihan Xu, Yuzheng Xu, Zhongxing Xu, Zhiling Yan, Boqin Yuan, Ruiqi Zhang, Yifan Zhang, Zibo Zhao, Liana, Santanu Bosu Antu, Haoyue Bai, Carlo Bosio, Joseph Cavanagh, Patricia Cavazos-Rehg, Tianxing Chen, Xuewen Chen, Yipu Chen, Zhu Chenyu, Chen Dai, Stefano De Castro, Yunfu Deng, Kaustubh Dhole, Jiayuan Ding, Chenchen Du, Zhehang Du, Hao Fan, Run-ze Fan, Hengyu Fu, Shi Gu, Yifan Gu, Charlie Guo, Baihe Huang, Baixiang Huang, Rimika Jaiswal, Zhihan Jiang, Ran Jin, Erin Kasson, Xin Lan, Joseph Lee, Deren Lei, Chenyu Li, Daofeng Li, Haitao Li, Hongwei Li, Jingyan Li, Xiao Li, Yi Li, Yinsheng Li, Yuangang Li, Zhixu Li, Wenyu Liang, Longtai Liao, Kevin Qinghong Lin, AndyZeyi Liu, Che Liu, Jiaming Liu, Kaiyuan Liu, Xuan Liu, Pan Lu, Wenbo Lv, Yicheng Lv, Qiuyang Mang, Kyle Montgomery, Yuzhou Nie, Ruoxi Ning, Jorin Overwiening, Xu Pan, Layna Paraboschi, Core Francisco Park, Justin Purnomo, Swati Rajwal, Scott Rankin, Bixuan Ren, Yiren Rong, HaoYang Shang, Ventus Shaw, Fiona Shen, Jiawei Shen, Minqi Shi, Qiu Shi, Huaxiu Yao, Tianneng Shi, Jonah So, Vladislav Susoy, Hannah Szlyk, Haocheng Wang, Jialu Wang, Wei Wang, Xinyu Wang, Zehao Wang, Dowling Wong, Angela Wu, Dehao Wu, Fangyu Wu, Mengyuan "Millie" Wu, Yu Wu, Yuchen Wu, Yuhao Wu, Qingpo Wuwu, Weihang Xiao, Yongyi Xiong, Fan Xu, Ruiling Xu, Mingxuan Yan, Benjamin Yang, Jirong Yang, Sen Yang, Xiaoli Yang, Yushi Yang, Haoran Ye, Xiaohu Yu, Zhengming Yu, Chenlong Zhang, Chi Zhang, Hanning Zhang, Hanwen Zhang, Junge Zhang, Kunpeng Zhang, Song Zhang, Wenjin Zhang, Wenshuo Zhang, Ying Zhang, Yizhi Zhang, Brian Zhao, Qijian Zhao, Yimin Zhao, Yuhaohua Zheng, Liwei Zhou, Tianyue Zhou, Sichen Zhu, Siqi Zhu, Yan Zhu, Yishu Zhu, Jierui Zuo, Chonghao Cai, Helena Casademunt, Wenjia Chen, Benjamin Cheng, Nawen Deng, Rao Fu, Tianfu Fu, Yifan Han, Ren He, Zhenyu He, Qiao Jin, Lang Lang, Yuetai Li, Sylvia Liu, Lu Lu, Qing Lu, Subhabrata Mukherjee, Yunqi Ouyang, Yin Ren, Dawei Shi, Haoran Wu, Zhiyue Wu, Hannah Yao, Zhuoran Yi, Jenny Yu, Rhea Zhan, Hang Zhou, Blake Zhu, Junfan Zhu, Alan Yuille, Yang Liu, Russell Alan Poldrack, Jiachen Li, Zhenglu Li, Molei Tao, Jing Huang, Wenqi Shi, Costas Spanos, Lichao Sun, Chenguang Wang, Orson Xu, Zhen Dong, Hector Gomez, Aylin Caliskan, Ali Emami, Haimin Hu, Zhi Li, Lihui Liu, Murphy Niu, Yi Shao, Jianxin Sun, Mikko Tolonen, Ting Wang, Sanjiv Das, Yanjun Gao, Wenbo Guo, Erika J Schneider, Zhiyong Lu, Mark Mueller, Radha Poovendran, Somayeh Sojoudi, Dawn Song</p><p><b>Upvotes:</b> 66</p><p><b>Summary:</b> Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional domains. We argue that this gap is largely an evaluation problem: widely used benchmarks lack sustained performance measurement on real and economically valuable workflows. This paper introduces Agents' Last Exam (ALE), a benchmark designed to evaluate AI agents on long-horizon, economically valuable, real-world tasks with verifiable outcomes. Developed in collaboration with 250+ industry experts, ALE covers non-physical industries defined with reference to O*NET / SOC 2018 (the U.S. federal occupational taxonomy). It is organized around a task taxonomy with 55 subfields grouped into 13 industry clusters covering 1K+ tasks. Current results show that the hardest tier remains far from saturated: across mainstream harness and backbone configurations, the average full pass rate is 2.6%. ALE is designed as a living benchmark: its task pool grows continuously as new workflows and industries are onboarded. More broadly, ALE is intended not merely as another leaderboard, but as an instrument for closing the gap between benchmark success and GDP-relevant impact.</p> + https://arxiv.org/abs/2606.05405 Wed, 03 Jun 2026 00:00:00 +0000 - WorldBench: A Challenging and Visually Diverse Multimodal Reasoning Benchmark - https://arxiv.org/abs/2606.06538 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06538.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yida Yin, Harish Krishnakumar, Chung Peng Lee, Boya Zeng, Wenhao Chai, Shengbang Tong, Wenhu Chen, Hu Xu, Xingyu Fu, Gabriel Sarch, Aleksandra Korolova, Zhuang Liu</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> In real-world applications, models are expected to perform reliably across diverse settings. Yet, many existing multimodal benchmarks expand task types without capturing the visual diversity needed to handle open-ended visual inputs. We present WorldBench, a challenging and visually diverse reasoning benchmark to evaluate Multimodal Large Language Models (MLLMs). We build a taxonomy of thousands of visual concepts across multiple domains (e.g., living things). Guided by this taxonomy, we curate a broad collection of images from search engines and existing datasets to comprehensively represent the visual world. Through structured trial-and-error, we manually design challenging questions that frontier MLLMs fail to answer. On quantitative and human evaluations, WorldBench achieves higher visual diversity than any existing diverse benchmark. Evaluating 15 MLLMs on WorldBench reveals weaknesses in visual understanding: even the strongest model reaches only 64.0% accuracy, while some models perform marginally above chance-level. We hope our work highlights the importance of visual diversity in building multimodal benchmarks.</p> - https://arxiv.org/abs/2606.06538 + Cosine Misleads: Auxiliary Losses Reshape Vision Language Models, Not Their Latents + https://arxiv.org/abs/2606.05753 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05753.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> XiuYu Zhang, Junfeng Fang, Zhenkai Liang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Latent visual reasoning (LVR) inserts supervised latent tokens between perception and answer generation in vision-language models (VLMs). The field uses alignment between these latents and their visual targets, i.e., cosine similarity or mean squared error (MSE), as both the training loss and the quality metric, assuming that better alignment yields a better answer. We test this with a designed matrix of five LVR variants and find the assumption inverted: cosine alignment is negatively correlated with accuracy across all five (r=-0.94). To explain this, we introduce PRISM, a pair of inference-time diagnostics: a linear probe that asks where the answer is decodable, and a corruption test that asks whether the latent is load-bearing. The supervised latents are largely bypassed. Corrupting them shifts accuracy by at most four points. The answer is decodable downstream of the latent but not at it, and the size of this decodability gap predicts how much each variant relies on its latent under perturbation. Consistent with an Information Bottleneck reading of the loss, the auxiliary objective reshapes the language model via shared parameters rather than via the latent variable it nominally optimizes.</p> + https://arxiv.org/abs/2606.05753 Thu, 04 Jun 2026 00:00:00 +0000 - Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation - https://arxiv.org/abs/2606.05988 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05988.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Maxime Griot, Paul Steven Scotti, Tanishq Mathew Abraham</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Reasoning models produce long chain-of-thought traces that are costly to distill and encourage verbose student outputs. We study post-hoc compression of such traces before knowledge distillation. Two teachers, Qwen3.5-397B-A17B and gpt-oss-120B, generate about 283k correct traces each; two instruction-tuned models then compress them to 8.6-21.0% of their original character length. Across a 48-run main grid plus seven Qwen-teacher truncation ablations, compressed traces reduce training tokens to 12-30% of raw, speed up training by 2.0-7.6x, and shorten inference outputs by 3-19x with smaller reductions under the shorter gpt-oss teacher. However, raw traces retain the highest downstream accuracy at every scale and for both teachers. A length-matched raw-trace truncation ablation shows that compression is not merely benefiting from a smaller token budget: model-compressed traces usually beat or match naive truncation, especially for smaller students, while maintaining shorter inference outputs. Overall, reasoning-trace compression offers an accuracy-efficiency trade-off rather than a free improvement: students retain up to 96% of raw-trace accuracy while gaining up to 18x higher per-token efficiency, and at the 0.8B scale under LoRA compressed traces narrow the raw-vs-compressed gap but do not exceed raw.</p> - https://arxiv.org/abs/2606.05988 + AsyncWebRL: Efficient Multi-Step RL for Visual Web Agents + https://arxiv.org/abs/2606.05597 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05597.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hao Bai, Rui Yang, Chenlu Ye, Spencer Whitehead, Aviral Kumar, Tong Zhang</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Training vision-language web agents with multi-step RL is compute-intensive, with two dominant forms of inefficiency: idle GPUs in synchronous RL, and trajectories that use more steps and tokens than necessary. We present AsyncWebRL, which addresses both. On the system side, an asynchronous design overlaps rollout, gradient update, and policy refresh across iterations, paired with two web-agent-specific adaptations, namely an everlasting rollout pool and lightweight screenshot handling, that together deliver up to a 2.9times end-to-end training-throughput speedup over the previously fastest open synchronous pipeline (WebGym). On the algorithmic side, we identify the per-trajectory normalizer 1/|τ_i| in multi-step GRPO as the root cause of trajectory-level and token-level inefficiency: because failures are systematically longer than successes, it down-weights the negative gradient on failed tokens, so the policy keeps producing verbose memory schemas. Replacing 1/|τ_i| with a constant 1/k breaks this coupling, contracting trajectories while preserving aggregate success. Together, these contributions set a new open-source state of the art on the WebGym out-of-distribution test split (+5.8% relative over the 42.9% prior best), with the largest gains on the harder slices (+42% relative on Medium, +48% relative on Hard).</p> + https://arxiv.org/abs/2606.05597 Thu, 04 Jun 2026 00:00:00 +0000 - Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation - https://arxiv.org/abs/2606.06712 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06712.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xingyu Su, Jacob Helwig, Shubham Parashar, Atharv Chagi, Lakshmi Jotsna, Degui Zhi, James Caverlee, Dileep Kalathil, Shuiwang Ji</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> We study the transformation of autoregressive models (ARLMs) into diffusion language models (DLMs). Rather than pretraining from scratch, prior work replaces the causal attention in ARLMs with bidirectional attention and then trains the resulting model using a DLM objective. However, these approaches incur two distribution shifts. First, transitioning from a next-token prediction objective to a DLM objective can discard knowledge acquired by the ARLM during training. Second, standard DLMs suffer from a train-inference mismatch, as the training loss is defined on randomly masked sequences rather than the trajectories encountered at inference produced by confidence-based decoding. To address both challenges, we introduce an On-Policy Diffusion Language Model (OPDLM) in which On-Policy Distillation (OPD) is employed for ARLM-to-DLM transformation. Specifically, OPDLM is trained via self-OPD, where the student, an ARLM with bidirectional attention, generates its own trajectories, and the teacher, the original frozen ARLM, distills its knowledge by providing target logits on these trajectories. By training directly in an on-policy manner, OPDLM eliminates the train-inference mismatch in DLMs, while distillation from the original model enhances knowledge retention from the ARLM. Empirical results demonstrate that OPDLM requires 15x to 7,000x fewer training tokens with strong performance across a wide variety of tasks. OPDLM avoids the prohibitive cost of DLM pretraining and positions DLM transformation as a form of ARLM post-training.</p> - https://arxiv.org/abs/2606.06712 + A Geometric Account of Activation Steering through Angle-Norm Decomposition + https://arxiv.org/abs/2606.06735 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06735.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Georgii Aparin, Tatiana Gaintseva</p><p><b>Upvotes:</b> 15</p><p><b>Summary:</b> Linear activation steering has gained popularity as a simple and empirically effective way to control language model behavior. More recently, spherical steering paradigms have been proposed to address limitations of additive interventions, often motivated by the assumption that hidden-state norm does not carry concept-relevant information. In this work, we revisit this assumption through a controlled empirical study designed to disentangle the roles of angular and radial components. We show that steering methods differ mainly in how they couple two geometric effects: changing a token's angular alignment with a concept direction and changing its hidden-state norm. Across seven language models, we find that concepts are represented primarily in angular structure, supporting the motivation for spherical methods, but that norm remains important for the stability and downstream effects of steering. Our results explain why interventions with similar concept-level effects can behave differently, and suggest that activation steering should be parameterized by interpretable angular and radial components of the intervention, rather than by a single additive coefficient that entangles these two effects.</p> + https://arxiv.org/abs/2606.06735 Thu, 04 Jun 2026 00:00:00 +0000 - Almieyar-Oryx-BloomBench: A Bilingual Multimodal Benchmark for Cognitively Informed Evaluation of Vision-Language Models - https://arxiv.org/abs/2606.05531 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05531.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mohammad Mahdi Abootorabi, Omid Ghahroodi, Anas Madkoor, Marzia Nouri, Doratossadat Dastgheib, Mohamed Hefeeda, Ehsaneddin Asgari</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Despite the rapid progress of Vision-Language Models (VLMs), the field lacks benchmarks that rigorously diagnose their true reasoning abilities and chart meaningful progress toward human-like multimodal intelligence. Most existing evaluations focus on piecemeal or disconnected tasks, obscuring critical cognitive weaknesses and providing little insight for targeted improvement. To address this gap, we introduce BloomBench, part of the Almieyar benchmarking series, the first cognitively human-grounded, bilingual (English-Arabic) multimodal benchmark for VLMs. Grounded in Bloom's Taxonomy, BloomBench systematically evaluates six levels of cognition (Remember, Understand, Apply, Analyze, Evaluate, Create) through carefully designed image-question-answer tasks. Built with a semi-automated pipeline and validated through a stratified hybrid quality assurance protocol, it ensures scalability, cultural inclusivity, and linguistic fidelity. Leveraging this framework, we conduct a comprehensive study of state-of-the-art VLMs to diagnose their cognitive profiles. Our analysis reveals a sharp cognitive asymmetry: while state-of-the-art models achieve strong performance ceilings in semantic understanding, they struggle substantially with factual recall and creative synthesis. This demonstrates that current general multimodal proficiency masks deeper limitations in specific cognitive layers. Furthermore, our study highlights a critical performance gap between Arabic and English, exposing limitations in current cross-lingual multimodal reasoning. These findings establish a foundation for developing more cognitively aligned and inclusive VLMs. The benchmark framework and dataset is available at: https://github.com/qcri/Almieyar-Oryx-BloomBench.</p> - https://arxiv.org/abs/2606.05531 + LatentSkill: From In-Context Textual Skills to In-Weight Latent Skills for LLM Agents + https://arxiv.org/abs/2606.06087 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06087.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Aofan Yu, Chenyu Zhou, Tianyi Xu, Zihan Guo, Rong Shan, Zhihui Fu, Jun Wang, Weiwen Liu, Yong Yu, Weinan Zhang, Jianghao Lin</p><p><b>Upvotes:</b> 50</p><p><b>Summary:</b> Agent systems increasingly use textual skills to encode reusable task procedures, but injecting these skills into the prompt at every step incurs substantial context overhead and exposes skill content as plaintext. We present LatentSkill, a framework that converts textual skills into plug-and-play LoRA adapters through a pretrained hypernetwork. LatentSkill stores skill knowledge in weight space rather than context space, removing per-step skill tokens while preserving modular loading, scaling, and composition. On ALFWorld and Search-QA, LatentSkill outperforms the corresponding in-context skill baseline while using substantially fewer prefill tokens: it improves ALFWorld success by 21.4 and 13.4 points on the seen and unseen splits with 64.1% fewer prefill tokens, and improves Search-QA exact match by 3.0 points with 72.2% lower skill-token overhead. Further analysis shows that generated skill LoRAs form a structured semantic geometry, can be precisely controlled via the LoRA scaling coefficient, and can be composed through parameter-space arithmetic when skill components are aligned. These findings suggest that weight-space skills provide an efficient, modular, and less exposed substrate for extending LLM agents.</p> + https://arxiv.org/abs/2606.06087 Thu, 04 Jun 2026 00:00:00 +0000 - Physics in 2-Steps: Locking Motion Priors Before Visual Refinement Erases Them - https://arxiv.org/abs/2606.06361 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06361.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Woojung Han, Seil Kang, Youngjun Jun, Min-Hung Chen, Fu-En Yang, Seong Jae Hwang</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Image-to-Video diffusion models leverage input images to generate visually stunning content, yet frequently produce motion that violates physical laws. We reveal a surprising finding: a 2-step generation often exhibits better physical consistency than a 50-step output from the same model. Through spectral analysis, we trace this to phase erosion during denoising; the phase degrades significantly (dropping by approx 18% from step 2 to step 50), whereas the magnitude remains relatively stable. Building on this insight, we propose PhaseLock, a training-free framework that preserves the valid motion priors from few-step inference throughout the denoising trajectory. Rather than relying on full-step inference for physical consistency, PhaseLock extracts a motion prior from just 2 steps and enforces it onto high-fidelity generation via Latent Delta Guidance. Our approach effectively mitigates phase degradation, improving physical consistency by an average of 6.2 points across diverse models while largely maintaining visual fidelity, with negligible overhead (1.06times time, 1.02times memory) and reduced reliance on expensive external guidance methods (sim5times time).</p> - https://arxiv.org/abs/2606.06361 - Thu, 04 Jun 2026 00:00:00 +0000 + Answer Presence Drives RAG Rewriting Gains + https://arxiv.org/abs/2606.05633 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05633.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yuejie Li, Yueying Hua, Ke Yang, Li Zhang, Yueping He, Yueping He, Ruiqi Li, Bolin Chen, Tao Wang, Bowen Li, Chengjun Mao</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Retrieval-augmented QA pipelines often route retrieved passages through an LLM rewriter before a smaller reader, lifting F1 by tens of points on multi-hop benchmarks; this gain is typically credited to improved evidence quality. We ask whether that lift is causally driven by the gold answer string appearing in the rewritten context rather than by curation per se, using a controlled intervention audit. For each rewritten context we re-run the reader after one of four controlled edits to the compile output: removing the gold answer span, replacing a length-matched random non-answer span (placebo), or injecting the gold into rewrites where it was absent (at the prefix or at a midpoint sentence boundary). Across twelve completed (cell, baseline) intervention runs spanning three reader families (Qwen2.5-7B, Qwen3.5-35B, GLM-4.7), two datasets (HotpotQA, 2WikiMultihopQA), and three compiler arrangements (MA-only, MB-only, MA+verify), removing the gold answer drops reader F1 by 28 to 64 points beyond the length-matched placebo on paired answer-in-compile strata, and prepending the gold into rewrites that lacked it raises F1 by +0.7 to +9.7 points in 10 of 12 (cell, baseline) combinations. A companion five-sentinel audit shows the conventional single-[MASK] probe is itself sentinel-fragile: on 2Wiki it reports a +4.12~F1 ``non-leakage residual'' that flips to -3.33 to -7.81~F1 under four alternative sentinels and fails an equivalence test for three of those four (1/4~pass). We do not propose a new rewriter or mitigation; we release the intervention runner and the sentinel panel so that other rewriter-gain claims can be tested against the same standard.</p> + https://arxiv.org/abs/2606.05633 + Thu, 04 Jun 2026 03:00:42 +0000 - Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators - https://arxiv.org/abs/2606.06476 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06476.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chenming Zhu, Jingli Lin, Yilin Long, Peizhou Cao, Tai Wang, Jiangmiao Pang, Xihui Liu</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> While Vision-Language Models (VLMs) have shown strong visual reasoning capabilities, their spatial reasoning abilities remain largely constrained to the observed images and text-oriented chain-of-thought. They often struggle to infer unobserved layouts, maintain cross-view consistency, and reason from alternative viewpoints when only limited egocentric observations are available. In this work, we study this problem as thinking with imagination, where a VLM actively acquires imagined visual evidence by interacting with a world simulator during reasoning. We propose Astra, an agentic spatial reasoning framework that empowers VLMs with action-conditioned visual imagination. Specifically, Astra couples Astra-VL, an RL-trained VLM policy, with Astra-WM, a Bagel-based world simulator that generates novel-view observations from context images and natural-language camera motions. To provide reliable imagined evidence, Astra-WM is trained with view consistency tuning to improve pose and content consistency across views. In the RL stage, we propose a world-simulator-in-the-loop two-phase RL curriculum to stabilize tool-use exploration and advance the model's ability to invoke the simulator only when imagined observations improve over direct answering. Experiments demonstrate that both the world simulator and the agentic policy are necessary: Astra-WM improves simulator-augmented Gemini-3-Flash on MMSI-Bench from 45.1 to 49.5, while Astra-VL improves the Qwen3-VL backbone from 29.8 to 38.8 on MMSI-Bench and from 36.8 to 42.7 on MindCube. These results show that imagined observations can provide useful spatial evidence, but effective world-model-augmented reasoning requires learning when, where, and how to imagine.</p> - https://arxiv.org/abs/2606.06476 - Thu, 04 Jun 2026 00:00:00 +0000 - - - LLM Explainability with Counterfactual Chains and Causal Graphs - https://arxiv.org/abs/2606.05972 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05972.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Nirit Nussbaum-Hoffer, Nitay Calderon, Liat Ein-Dor, Roi Reichart</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Causal graphs provide a high-level language for making mechanisms transparent. Recent work uses Large Language Models (LLMs) to recover causal graphs of external-world processes. Instead, in this paper, we use causal graphs to model LLM inference itself, providing stakeholders with a transparent view of how the model perceives and organizes high-level concepts to produce a prediction. We propose a four-phase method for constructing such graphs. Given a target LLM and a set of textual examples, our method discovers class-discriminative, human-interpretable concepts and maps each input to LLM-perceived concept states. We then introduce an MCMC-inspired counterfactual augmentation procedure that expands the sparse observational data through chains of counterfactuals. This enables stable causal discovery with σ-CG, yielding informative, interpretable graphs. We apply our method to three LLMs across disease diagnosis, sentiment analysis, and LLM-as-a-judge classification tasks. We evaluate the learned graphs for predictive fidelity and structural stability, and the MCMC-inspired augmentation for convergence and downstream utility. Our results show that the discovered causal graphs capture meaningful dependencies consistent with LLMs' reasoning. Together, this paper provides a foundation for concept-level explainability of LLMs.</p> - https://arxiv.org/abs/2606.05972 - Thu, 04 Jun 2026 00:00:00 +0000 - - - UnpredictaBench: A Benchmark for Evaluating Distributional Randomness in LLMs - https://arxiv.org/abs/2606.06622 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06622.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Amirhossein Abaskohi, Amirhossein Dabiriaghdam, Liang Luo, Ellie Dingqiao Wen, Lele Wang, Giuseppe Carenini, Peter West</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> We introduce UnpredictaBench, an evaluation that tests the ability of large language models (LLMs) to capture true underlying distributions. As LLMs are increasingly used as substitutes for other entities (e.g., for humans in economic simulations), the tendency of many models to collapse towards a single plausible answer means a failure to capture the unpredictability of real systems. Recent work on improving output diversity is insufficient for this setting: simulation requires samples that are calibrated to a target distribution, not merely varied outputs. UnpredictaBench isolates a simplified but fundamental version of this problem: sampling outcomes from individual target distributions, including canonical statistical distributions, distributions induced by stochastic programs, and natural-language scenarios that describe random processes. We introduce 448 such problems together with KS@N, a general-purpose evaluation metric that quantifies how well a model outputs approximate black-box target distributions via the Kolmogorov-Smirnov statistical test. This is the rate at which we fail to reject model samples of size N against ground-truth samples, with larger N indicating greater difficulty. Tested across open and proprietary models, we find a large spread in distributional capabilities. For instance, when models generate samples of size 100 (KS@100, our standard metric), scores range from near 0 to over 20%. No model is able to achieve over 40% at KS@100, showing significant headroom in distributional sampling as a capability. Although adding reasoning can somewhat increase scores, we find no immediate solution for this issue. UnpredictaBench shows that even simple distributional simulation remains challenging, making it a necessary first step toward using LLMs as stand-ins for complex systems.</p> - https://arxiv.org/abs/2606.06622 - Thu, 04 Jun 2026 00:00:00 +0000 - - - SubtleMemory: A Benchmark for Fine-Grained Relational Memory Discrimination in Long-Horizon AI Agents - https://arxiv.org/abs/2606.05761 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05761.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wenxuan Wang, Haoyu Sun, Fukuan Hou, Mingyang Song, Weinan Zhang, Yu Cheng, Yang Yang</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Persistent AI assistants, such as OpenClaw, accumulate large collections of related memories over long-term interactions. As these memories grow, they may reinforce one another, diverge across contexts, or directly conflict, making correct assistance depend on memory relations rather than isolated recall. Existing long-term memory benchmarks rarely probe how agents preserve and utilize such relations during downstream tasks. To address this gap, we introduce SubtleMemory, a benchmark for fine-grained relational memory discrimination in long-running AI agents. SubtleMemory constructs relation-controlled latent semantic artifacts whose variants instantiate complementary, nuanced, or contradictory relations, and embeds them into realistic user-agent histories, requiring agents to recover distributed relational structures during later queries and instructions. The benchmark contains 1,522 evaluation instances over 10 long histories, grounded in 1,090 relation-controlled memory-variant sets and spanning user-related and non-user-related queries. Evaluating six standalone memory systems, two Claw-style agents with native memory modules, and three Claw-style agents with plugin memory modules, we find that current systems remain weak on fine-grained relational memory discrimination. We further introduce diagnostic protocols that reveal distinct capability profiles across memory preservation, retrieval, and downstream reasoning stages.</p> - https://arxiv.org/abs/2606.05761 - Thu, 04 Jun 2026 00:00:00 +0000 - - - When Tools Fail: Benchmarking Dynamic Replanning and Anomaly Recovery in LLM Agents - https://arxiv.org/abs/2606.05806 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05806.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Dongsheng Zhu, Xuchen Ma, Yucheng Shen, Xiang Li, Yukun Zhao, Shuaiqiang Wang, Lingyong Yan, Dawei Yin</p><p><b>Upvotes:</b> 18</p><p><b>Summary:</b> Existing benchmarks evaluate Tool-Integrated Reasoning (TIR) in LLMs on idealized ''happy paths'', largely overlooking real-world tool failures. We introduce ToolMaze, a benchmark for dynamic path discovery and error recovery in TIR agents. To separate systematic replanning from blind trial-and-error, ToolMaze adopts a two-dimensional design: DAG-based topological complexity and a 2 times 2 taxonomy of tool perturbations (explicit/implicit, transient/permanent). Evaluations show that perturbations degrade performance across nearly all models, with the sharpest drops under implicit semantic failures. Driven by systemic over-trust in corrupted outputs, Perturbation Recovery Rate (PRR) plummets by around 37\% in these scenarios, while complex topologies trap agents in futile trial-and-error loops. Crucially, agentic fault-tolerance improves with model scale 3.66times slower than basic task execution, highlighting dynamic replanning as a distinct bottleneck unaddressed by model scaling or prompting. Data and code are available at https://github.com/Zhudongsheng75/ToolMaze.</p> - https://arxiv.org/abs/2606.05806 - Thu, 04 Jun 2026 00:00:00 +0000 - - - OpenSkill: Open-World Self-Evolution for LLM Agents - https://arxiv.org/abs/2606.06741 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06741.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhiling Yan, Dingjie Song, Hanrong Zhang, Wei Liang, Yuxuan Zhang, Yutong Dai, Lifang He, Philip S. Yu, Ran Xu, Xiang Li, Lichao Sun</p><p><b>Upvotes:</b> 18</p><p><b>Summary:</b> Self-evolving agents requires adaptation after deployment, but existing approaches assume a usable learning loop, such as curated skills, successful trajectories, or verifier signals. Real open-world deployments may provide none of these, offering only a task prompt. In this work, we study open-world self-evolution, where an agent must build both its skills and its own verification signals from scratch, using open-world resources but no target-task supervision. We propose OpenSkill, a framework that bootstraps this loop: it acquires grounded knowledge and verification anchors from documentation, repositories, and the web, synthesizes them into transferable skills, and refines those skills against self-built virtual tasks grounded in the anchors rather than in target answers. The open world thus supplies both the knowledge to be learned and a supervision-independent practice environment, with target-task supervision reserved for final evaluation. Across three benchmarks and two target agents, OpenSkill attains the best automated pass rate while satisfying the no-supervision constraint. Analysis shows its skills transfer across models without model-specific adaptation, and its self-built verifier aligns with ground-truth outcomes despite never accessing them.</p> - https://arxiv.org/abs/2606.06741 - Thu, 04 Jun 2026 00:00:00 +0000 - - - Robots Need More than VLA and World Models - https://arxiv.org/abs/2606.06556 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06556.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Elis Karcini, Faisal Mehrban, Quang Nguyen, Mac Schwager, Arash Ajoudani, Cesar Cadena, Jan Peters, Marco Hutter, Haitham Bou-Ammar</p><p><b>Upvotes:</b> 20</p><p><b>Summary:</b> Generalist robot intelligence is often framed as a policy-scaling problem: collect more robot demonstrations, train larger Vision-Language-Action (VLA) models, and expect broader generalisation. In this position paper, we argue that this framing is incomplete. The central bottleneck is not only policy learning, but the absence of mechanisms that convert the world's abundant unstructured behavioural data into grounded robot supervision. Human motion, internet video, simulation rollouts, and interactive demonstrations contain rich information about tasks, goals, contacts, failures, and physical constraints, yet most of this information is not directly usable by robot policies because it lacks embodiment-specific action labels, task semantics, and reward structure. We identify four missing components for the next generation of robotics: data interfaces for autolabelling unstructured behaviour, embodiment interfaces for retargeting human motion to robot actions, world-model interfaces for physics-grounded 3D reasoning, and reward interfaces for inferring task progress and success from video and language. We survey recent progress in robot foundation models, cross-embodiment datasets, learning from video, world models, and reward modelling, and propose a research agenda for building robotics systems that can learn not only from robot demonstrations, but from the broader physical world.</p> - https://arxiv.org/abs/2606.06556 - Thu, 04 Jun 2026 00:00:00 +0000 - - - Direct 3D-Aware Object Insertion via Decomposed Visual Proxies - https://arxiv.org/abs/2606.06601 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06601.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jingbo Gong, Yikai Wang, Yushi Lan, Yuhao Wan, Ziheng Ouyang, Rui Zhao, Ming-Ming Cheng, Qibin Hou, Chen Change Loy</p><p><b>Upvotes:</b> 22</p><p><b>Summary:</b> Object insertion aims to seamlessly composite a reference object into a specified region of a background image. Recent diffusion-based methods achieve high visual quality but formulate insertion as a simple 2D inpainting task, providing no explicit control over the object's 3D pose and limiting their practical applicability. We propose DIRECT (Decomposed Injection for Reference Composition and Target-integration), a novel framework that integrates interactive pose manipulation with high-fidelity 2D image synthesis to enable pose-controllable object insertion. Our method decomposes the insertion conditions into three complementary components: appearance guidance capturing visual details from the reference object, geometry guidance derived from the user-adjusted 3D proxy, and context guidance from the target background. By injecting them through separate pathways, DIRECT avoids feature entanglement and simultaneously preserves reference appearance, follows the user-specified pose, and adapts the object to the target scene. We also introduce an automated data construction pipeline to improve the diversity and quality of training data. Experiments show that DIRECT outperforms previous methods in both geometric controllability and visual quality.</p> - https://arxiv.org/abs/2606.06601 - Thu, 04 Jun 2026 00:00:00 +0000 - - - SoCRATES: Towards Reliable Automated Evaluation of Proactive LLM Mediation across Domains and Socio-cognitive Variations - https://arxiv.org/abs/2606.05563 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.05563.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Taewon Yun, Hyeonseong Park, Jeonghwan Choi, Hayoon Park, Yeeun Choi, Hwanjun Song</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> Evaluating LLM mediators remains challenging, as mediation unfolds as a real-time trajectory shaped by disputants' shifting emotions, intentions, and context. Existing testbeds rely on a few expert-authored domains, vary mainly strategic posture, and score every turn against every topic, introducing off-topic noise. We introduce SoCRATES, a benchmark for evaluating proactive LLM mediators in realistic, multi-domain testbeds. It constructs scenarios from real conflicts through an agentic pipeline across eight domains, probes five socio-cognitive adaptation axes (strategic posture, party composition, history length, emotional reactivity, and cultural identity), and scores each topic only on the turns that advance it via a topic-localized evaluator. The evaluator reaches 0.82 alignment with human experts, more than doubling a per-turn baseline. Benchmarking eight frontier LLMs, we find that even the strongest mediator closes only about a third of the unmediated consensus gap under diverse and realistic testbeds, with performance varying sharply by socio-cognitive axis, highlighting that progress lies in social adaptation to diverse conditions.</p> - https://arxiv.org/abs/2606.05563 - Thu, 04 Jun 2026 00:00:00 +0000 - - - How Far Can Chord-Symbol Time-Series Adaptation Carry Genre Identity? Capabilities and Boundaries in Multi-Genre Chord-Symbol Modeling - https://arxiv.org/abs/2606.07334 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07334.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jinju Lee</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Harmony is a compact symbolic layer where mathematical pitch relations, acoustic consonance, and musical convention meet. This report treats chord-symbol sequences not as a complete representation of music, but as an interpretable, controllable time series for genre-local harmonic modeling. Starting from a frozen pop-jazz Music Transformer checkpoint, I evaluate how far small adaptation interfaces can extend the model to eleven target genres: blues, bossa nova, Bach chorales, country, electronic, folk, funk, gospel, hip-hop, R&B/soul, and rock. The main evaluation compares LoRA, IA3, BitFit, prefix tuning, and full fine-tuning over 11 genres and 3 seeds, a complete 165-cell grid. All five methods improve over the frozen base on held-out chord prediction, with macro gains from +2.89 to +3.61 points; LoRA and IA3 score highest, but Wilcoxon tests with Holm and Benjamini-Hochberg correction do not support a decisive winner. A matched-data-size control sharpens this: when genres are sub-sampled to a common corpus size, IA3 stays on top but LoRA's full-data edge disappears and it falls to last, indicating the small gaps are partly data-driven. A control-token baseline is also strong, and wrong-genre adapters often beat the frozen base, suggesting much of the effect comes from lightweight conditioning over a reusable harmonic base rather than one particular adapter family. Additional diagnostics (rank sweeps, wrong-genre rotation, a base-checkpoint ablation, chord-only genre classification, generated-output statistics, real-song evaluation, and duplicate analysis) support a bounded conclusion: chord-symbol adaptation reliably improves genre-local harmonic prediction, but chord symbols alone do not carry complete genre identity. The report therefore avoids claims about perceived genre authenticity or full musical quality, which require controlled listener or musician evaluation.</p> - https://arxiv.org/abs/2606.07334 + Robotic Policy Adaptation via Weight-Space Meta-Learning + https://arxiv.org/abs/2606.07217 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07217.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Christian Bianchi, Siamak Yousefi, Alessio Sampieri, Andrea Roberti, Luca Rigazio, Fabio Galasso, Luca Franco</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Vision-Language-Action (VLA) models are emerging as a promising paradigm for robotic manipulation, enabling general-purpose policies trained from large corpora of demonstrations and action labels. However, adapting these models to new tasks still typically requires task-specific demonstrations, action annotations, and additional fine-tuning, making deployment costly and difficult to scale. We propose WIZARD, a weight-space meta-learning framework that sidesteps task-specific fine-tuning by generating task-specific LoRA parameters for a frozen VLA policy. Given only a language instruction and a short demonstration video, WIZARD predicts the corresponding adaptation weights in a single forward pass, without target-task action labels or test-time optimization. During meta-training, WIZARD learns to map task evidence directly to expert LoRA updates, capturing relationships between tasks in weight space. Experiments on LIBERO show that WIZARD improves performance by up to ~2x on unseen dataset collections and up to ~14x on unseen tasks. On a Franka Emika Panda, WIZARD consistently improves over a real-domain adapted baseline, showing that generated adapters provide task-level specialization beyond simulation.</p> + https://arxiv.org/abs/2606.07217 Fri, 05 Jun 2026 00:00:00 +0000 - Empirical Study on the Characteristics and Evolution of AI-usage in GitHub Repositories: Evidence from Code Comments - https://arxiv.org/abs/2606.06843 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06843.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Abdullah Al Mujahid, Preetha Chatterjee, Mia Mohammad Imran</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Developers increasingly use AI tools such as ChatGPT, Copilot, and Claude in everyday software workflows, but prior studies often evaluate LLM outputs in isolation rather than examining how developers adapt them in real projects. We analyze 35,361 GitHub code comments that explicitly reference AI use and their associated code blocks. We first open-code 500 unique comments and code blocks to derive a taxonomy of AI-assisted development activities, then annotate the full dataset using two LLM-based classifiers and aggregate predictions with Dawid-Skene expectation-maximization. We also analyze 12,996 subsequent commit messages to study how AI-assisted code evolves after introduction, and examine temporal trends from December 2022 to March 2026. Our results show that developers primarily use LLMs for code implementation, followed by code enhancement, debugging, documentation, and testing. Subsequent commits frequently involve refactoring and cleanup, feature integration and extension, and bug fixing, indicating sustained human oversight in adapting AI-assisted code. Over time, AI-referencing comments shift from direct code generation toward knowledge and conceptual support and code enhancement. These findings suggest that AI tools are becoming embedded not only as code-generation aids, but also as collaborative support mechanisms whose outputs are refined, extended, and corrected by developers over time.</p> - https://arxiv.org/abs/2606.06843 + DuMate-DeepResearch: An Auditable Multi-Agent System with Recursive Search and Rubric-Grounded Reasoning + https://arxiv.org/abs/2606.07299 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07299.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Lingyong Yan, Can Xu, Yukun Zhao, Wenxuan Li, Qingyang Chen, Jiulong Wu, Wenli Song, Xiangnan Li, Weixian Shi, Yiqun Chen, Xuchen Ma, Yuchen Li, Jiashu Zhao, Shuaiqiang Wang, Jianmin Wu, Dawei Yin</p><p><b>Upvotes:</b> 5</p><p><b>Summary:</b> Deep Research (DR) has emerged as a new agentic paradigm to tackle complex, open-ended research tasks, demanding systems that can iteratively frame problems, acquire evidence, verify sources, and synthesize long-form reports. In practice, however, current DR systems are constrained by four interrelated limitations: long-horizon planning over an underspecified scope, the bottleneck of decomposing and scheduling such tasks within a single agent, hallucination risk in long-form synthesis, and limited process auditability. This technical report presents DuMate-DeepResearch, a multi-agent DR framework built on the Qianfan Agent Foundry. The framework decouples the Agent Core, which handles task understanding, planning, and scheduling, from an extensible Tool Ecosystem for retrieval, evidence acquisition, and report rendering, making every intermediate decision and tool invocation explicitly traceable. Building on this infrastructure, DuMate-DeepResearch further introduces three mechanisms: (i) a graph-based dynamic planning strategy expands the research roadmap coarse-to-fine and continuously revises it through reflection, re-planning, backtracking, and parallel branching; (ii) a recursive two-level execution design delegates each complex search sub-task to an inner Search Agent that runs its own planning loop, isolating noisy retrieval and stabilizing long-horizon execution; (iii) a rubric-based test-time optimization mechanism dynamically generates task-specific quality criteria and uses them as live reasoning scaffolds for evidence-grounded synthesis and adaptive stopping. Across two deep research benchmarks, DuMate-DeepResearch establishes new state-of-the-art results: the best overall score (58.03%) on DeepResearch Bench, and the best overall score (61.95%) on DeepResearch Bench II while ranking first in information recall and analysis.</p> + https://arxiv.org/abs/2606.07299 Fri, 05 Jun 2026 00:00:00 +0000 - ECI_{sem}: Semantic Residual Effective Contrastive Information for Evaluating Hard Negatives - https://arxiv.org/abs/2603.20990 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2603.20990.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Aarush Sinha, Rahul Seetharaman, Aman Bansal</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Hard-negative source selection for dense retrieval is usually decided only after fine-tuning and downstream evaluation. We propose ECI_{sem}, a semantic residual variant of Effective Contrastive Information (ECI) that ranks candidate negative sources using frozen target-encoder embeddings. ECI_{sem} is training-free, not label-free: each scored example requires a query, a labeled positive, and an explicit candidate negative. ECI_{sem} builds a weighted residual information matrix from target consistency, semantic locality, lexical residuality, and a log-determinant diversity objective. On MS MARCO negative sources, in-family ECI_{sem} ranks LLM negatives highest among non-hybrid sources and Dense+LLM highest among hybrid sources, matching the strongest aggregate BEIR transfer results across DistilBERT, E5-base, and Contriever. Controlled ablations show that this alignment depends on using the target encoder family, while additional ablations show stability under sample-size, temperature, tokenizer, and IDF-corpus perturbations. The theory gives a local linearized link to loss reduction, while the empirical study treats downstream evaluation as the final test.</p> - https://arxiv.org/abs/2603.20990 + SlimSearcher: Training Efficiency-Aware Web Agents via Adaptive Reward Gating + https://arxiv.org/abs/2606.07074 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07074.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zequn Xie, Junjie Wang, Dan Yang, Jie Feng, Yue Shen, Jian Wang, Jinjie Gu</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Deep research agents have demonstrated remarkable capabilities in complex information-seeking tasks, yet this power comes at a steep computational cost. Driven by accuracy-focused training paradigms, current models adopt brute-force strategies characterized by blind tool dependency and performative reasoning-generating long, redundant trajectories that are far from necessary for resolving these tasks, leading to wasteful tool calls and excessive token consumption. To overcome this efficiency trap, we propose SlimSearcher, a principled framework that pushes the Pareto frontier between accuracy and computational cost across both Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL). In the SFT stage, SlimSearcher employs Pareto-efficient filtration to distill trajectories that are both successful and economical, guiding the model toward inherently efficiency-aware search behaviors. During RL, we introduce Adaptive Reward Gating, a dynamic reward-shaping mechanism that evaluates relative tool and token efficiency within a sampled cohort. By cascading these adaptive efficiency metrics with a strict correctness gate, our approach effectively avoids the brevity bias associated with absolute penalties and mitigates reward hacking. Extensive experiments on long-horizon benchmarks, including GAIA, BrowseComp, and XBenchDeepSearch, demonstrate that SlimSearcher reduces average tool-call rounds by 17%-58% while maintaining or improving accuracy.</p> + https://arxiv.org/abs/2606.07074 Fri, 05 Jun 2026 00:00:00 +0000 - Streaming Video Generation with Streaming Force Control - https://arxiv.org/abs/2606.07508 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07508.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hanhui Wang, Yiming Xie, Haiwen Feng, Zhaoyang Lv, Shenlong Wang, Huaizu Jiang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> We introduce StreamForce, a streaming video generation framework that enables physically grounded control through continuous force inputs. Unlike prior video models that train separate models for different force types, assume fixed forces, or rely on non-causal processing, StreamForce is a causal and unified model that responds instantly and coherently to both local and global, time-varying forces. To achieve this, we design a unified force representation as a control signal and develop a distillation pipeline for force-controllable video generation. Our model combines autoregressive efficiency with force responsiveness, sustaining stable photometric and dynamic realism. StreamForce runs at up to 16.6 FPS on a single GPU, achieving state-of-the-art performance in both force adherence and motion realism. Project website: https://neu-vi.github.io/StreamForce/</p> - https://arxiv.org/abs/2606.07508 + Whisper Hallucination Detection and Mitigation via Hidden Representation Steering and Sparse AutoEncoders + https://arxiv.org/abs/2606.07473 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07473.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Georgii Aparin, Vadim Popov, Tasnima Sadekova, Assel Yermekova</p><p><b>Upvotes:</b> 11</p><p><b>Summary:</b> Whisper, a widely adopted ASR model, is known to suffer from hallucinations - coherent transcriptions generated for non-speech audio entirely disconnected from the input. We investigate whether hallucinations can be detected and mitigated through Whisper's internal representations. We extract audio encoder activations and evaluate two representation spaces: raw Whisper activations and Sparse AutoEncoder (SAE) latents. We show that both spaces encode linearly separable hallucination-related information, with discriminative power concentrated in a sparse feature subset and increasing toward deeper encoder layers. We propose two steering strategies: activation-space steering and SAE latent-space steering. SAE-based steering reduces hallucination rate from 72.63% to 14.11% for Whisper small and from 86.88% to 27.33% for Whisper large-v3 on the full non-speech test set, with small WER degradation on speech data, approaching the performance of fine-tuning-based methods.</p> + https://arxiv.org/abs/2606.07473 Fri, 05 Jun 2026 00:00:00 +0000 - Towards Retrieving Interaction Spaces for Agentic Search - https://arxiv.org/abs/2606.06880 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06880.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shengyao Zhuang, Yuansheng Ni, Hengxin Fun, Jimmy Lin, Xueguang Ma</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Retrieval for search agents is still inherited from non-agentic information retrieval: a retriever ranks the corpus and the agent reads a small set of returned documents. Recent direct corpus interaction (DCI) work shows that agents can instead interact with the raw corpus through shell tools such as grep and file reads. But unbounded interaction does not scale: every broad shell command is a scan over the whole corpus, and latency degrades sharply as the corpus grows. We argue that the role of retrieval for agentic search is not just to select documents that fit in the LLM context window, but to construct an interaction space: a bounded subset of the corpus the agent can explore with associated tools. Two design consequences follow. The space needs a boundary supplied by retrieval, and the objects within it should be processed for interaction. As a proof of concept, we propose RISE (Retrieving Interaction SpacE): we use BM25 to construct the interaction space; meanwhile, its documents are processed during indexing for shell-style navigation. On BrowseComp-Plus, RISE matches the pure-shell DCI baseline at 78% accuracy with gpt-5.4-mini at roughly one quarter of the per-query cost. At 1M documents, RISE-BM25 reaches 81% on gpt-5.4-mini, whereas DCI on gpt-5.4-nano degrades to 60% with 33 of 100 wall-clock failures.</p> - https://arxiv.org/abs/2606.06880 + On the Geometry of On-Policy Distillation + https://arxiv.org/abs/2606.07082 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07082.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zhennan Shen, Yanshu Li, Qingyu Yin, Chak Tou Leong, Zhilin Wang, Yanxu Chen, Rongduo Han, Sunbowen Lee, Yi R. Fung</p><p><b>Upvotes:</b> 58</p><p><b>Summary:</b> On-policy distillation (OPD) is increasingly used to improve large language model reasoning, but its training dynamics remain poorly understood. We characterize the trajectory of OPD updates in parameter space and compare it with supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR). A suite of parameter-space diagnostics consistently places OPD in a relaxed off-principal regime: compared with SFT, its updates affect fewer weights and avoid principal directions more strongly, while compared with RLVR, they remain less tightly constrained. Beyond this static localization, OPD exhibits subspace locking: its cumulative updates rapidly enter a narrow low-dimensional channel. Constraining training to the update subspace formed early in training preserves OPD performance but substantially degrades SFT, indicating that the locked subspace is functionally sufficient for OPD. Control experiments further show that sparsifying the update tokens and shifting rollout generation off-policy preserve the rank dynamics, whereas mixing the OPD objective with RLVR changes them. Overall, these results suggest that OPD is not merely an intermediate point between SFT and RLVR, but induces its own update geometry in parameter space.</p> + https://arxiv.org/abs/2606.07082 Fri, 05 Jun 2026 00:00:00 +0000 - Entropy as a Structural Prior: How a Log-Barrier on DiT Belief Space Drives Musical Diversity and Development - https://arxiv.org/abs/2606.07207 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07207.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zixi Li, Youzhen Li</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Confidence-based loss weighting is usually avoided in generative models because it accelerates errors when the model is confidently wrong, but this intuition breaks down in supervised diffusion training. We introduce the Eisbach log-barrier, a parameter-free weight derived from the entropy of the DiT output's spatial energy distribution: high entropy damps the gradient, while low entropy preserves it. Applied to LoRA fine-tuning of Stable Audio 3 Medium on MusicCaps, it unexpectedly yields stronger thematic development, clearer acoustic differentiation, and higher textural diversity than unweighted training, the opposite of mode collapse. This works because in supervised diffusion the gradient direction is locked to ground truth, so confidence only scales the step size, and because temporal entropy downweights flat samples while preserving high-contrast ones. The result is an online, self-referential data curriculum that emerges purely from the forward pass, with analyzed noise-level dynamics and testable predictions.</p> - https://arxiv.org/abs/2606.07207 + SWE-Explore: Benchmarking How Coding Agents Explore Repositories + https://arxiv.org/abs/2606.07297 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07297.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shaoqiu Zhang, Yuhang Wang, Jialiang Liang, Yuling Shi, Wenhao Zeng, Maoquan Wang, Shilin He, Ningyuan Xu, Siyu Ye, Kai Cai, Xiaodong Gu</p><p><b>Upvotes:</b> 96</p><p><b>Summary:</b> Repository-level coding benchmarks such as SWE-bench have driven a rapid surge in the capabilities of coding agents. Yet they usually treat coding tasks as a holistic, binary prediction problem (e.g., resolved or unresolved), neglecting fine-grained agent capabilities such as repository understanding, context retrieval, code localization, and bug diagnosis. In this paper, we introduce SWE-Explore, a benchmark that isolates the evaluation of repository exploration, a critical capability of coding agents. Given a repository and an issue, SWE-Explore asks an explorer to return a ranked list of relevant code regions under a fixed line budget. SWE-Explore covers 848 issues across 10 programming languages and 203 open-source repositories. For each instance, we derive line-level ground truth from independent agent trajectories that successfully solved the same issue, distilling the specific code regions their solution paths actually consulted. We evaluate exploration along coverage, ranking, and context-efficiency dimensions, showing that these metrics strongly track downstream repair behavior. Across a broad set of retrieval methods, general coding agents, and specialized localizers, we find that agentic explorers form a clear tier above classical retrieval. While file-level localization is already strong for modern methods, line-level coverage and efficient ranking remain the key axes differentiating state-of-the-art explorers.</p> + https://arxiv.org/abs/2606.07297 Fri, 05 Jun 2026 00:00:00 +0000 - Stream3D-VLM: Online 3D Spatial Understanding with Incremental Geometry Priors - https://arxiv.org/abs/2606.06891 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06891.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hanxun Yu, Xuan Qu, Lei Ke, Boqiang Zhang, Yuxin Wang, Jianke Zhu, Dong Yu</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> Despite advances in 3D scene understanding, existing 3D Large Multimodal Models operate in offline settings, requiring complete scene observations or predefined video clips. In this paper, we present an online 3D vision-language model that enables real-time spatial understanding from streaming video. Our approach adopts an autoregressive streaming control modeling based on the LLM's next-token prediction objective to learn when to respond, and employs a lightweight Visual-Spatial Feature Integration (VSFI) module to incrementally inject temporally aligned geometry priors into the visual stream. To alleviate long-context decoding overhead, we propose a plug-and-play Geometry-Adaptive Voxel Compression (GAVC) module for efficient visual token compression. To address the scarcity of streaming 3D-language data, we further develop a scalable data generation pipeline that curates over 1M online spatio-temporal 3D QA pairs and establishes a comprehensive benchmark spanning 29 tasks. Extensive experiments show that our approach significantly outperforms both proprietary and open-source models across online and offline 3D spatial understanding, reasoning, and grounding tasks. The project page is available at https://stream3d-vlm.github.io/</p> - https://arxiv.org/abs/2606.06891 - Fri, 05 Jun 2026 00:00:00 +0000 + SigmaScale: LLM Compression with SVD-based Low-Rank Decomposition and Learned Scaling Matrices + https://arxiv.org/abs/2606.07098 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07098.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ernests Lavrinovics, Marco Letizia, Roy Janco, Shai Segal, Johannes Bjerva, Maurizio Pierini</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> We present SigmaScale, a method for learning auxiliary scaling matrices S to aid truncated Singular Value Decomposition (SVD) based Large Language Model (LLM) compression. Instead of deriving scaling matrices analytically, SigmaScale optimizes two sets of vectors that define diagonal row and column scaling transformations under an activation-aware compression loss. We show that learned scaling lowers the effective intrinsic rank of weight matrices, as reflected by reductions in effective-rank entropy, and that this reduction is strongly correlated with compression loss. Experiments on Llama 3.1 8B Instruct and Qwen3-8B show that SigmaScale is competitive with closely related state-of-the-art SVD-based compression methods across perplexity and zero-shot benchmarks. By using learned activation-aware transformations, SigmaScale explores a more flexible route to low-rank LLM compression by adapting to the structure of individual model weights. The advantage observed in specific tasks makes our approach a valid option for applications requiring a reduced LLM-inference computing cost.</p> + https://arxiv.org/abs/2606.07098 + Fri, 05 Jun 2026 09:48:58 +0000 - Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills - https://arxiv.org/abs/2606.07412 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07412.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Chuan Xiao, Zhengbo Jiao, Shaobo Wang, Wei Wang, Bing Zhao, Hu Wei, Linfeng Zhang, Lin Qu</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> LLM-driven software engineering agents have become a central testbed for real-world language-model capability, yet their training remains limited by the availability of high-quality SWE tasks. Existing synthetic data methods typically create tasks through fixed mutation or bug-injection procedures, making the resulting distributions largely independent of the agent's own weaknesses and training progress. We introduce Socratic-SWE, a closed-loop self-evolution framework that reuses the agent's historical solving traces as a source of training signal. Rather than treating traces only as evidence for reward computation, Socratic-SWE distills them into structured agent skills that summarize recurring failures and effective repair patterns. These skills then guide the generation of targeted repair tasks in real repositories. Candidate tasks are checked through execution-based validation and scored with a solver-gradient alignment reward, so that the retained tasks are both verifiable and useful for improving the Solver. The updated Solver produces new traces, enabling the task curriculum to adapt over successive rounds. Across SWE-bench Verified, SWE-bench Lite, SWE-bench Pro, and Terminal-Bench 2.0, Socratic-SWE consistently improves over self-evolving baselines under the same compute budget, reaching 50.40% on SWE-bench Verified after three iterations. These results suggest that solving traces can serve as a scalable substrate for self-evolving SWE agents.</p> - https://arxiv.org/abs/2606.07412 - Fri, 05 Jun 2026 00:00:00 +0000 + Where Rectified Flows Leak: Characterising Membership Signals Along the Interpolation Path + https://arxiv.org/abs/2606.07271 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07271.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Thomas Sesmat, Gabriel Meseguer-Brocal, Geoffroy Peeters</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Understanding what generative models retain from training data remains challenging, with implications for copyright and privacy. Beyond verbatim reproduction, models can encode subtler traces of their training data that never surface in their outputs yet remain exploitable. We study this regime for Rectified Flows, which are increasingly used in deployed generative systems. We analyse the interpolation path X_λ= (1-λ)X_0 + λX_1 that defines the Rectified Flow training. We show that a gap exists between the reconstruction of train and test data that follows a bell-shaped curve over λ, wich accumulates during training, while the validation metrics remain stable. The signal has a maximum whose location we derive in closed form under Gaussian assumptions. We validate these predictions on both audio and images and show that the bell-shaped structure is universal, while the peak prediction holds when our assumptions are satisfied. As a proof of concept, we exploit this specific λ-resolved structure to perform a Membership Inference Attack, distinguishing members of the training set from non-members.</p> + https://arxiv.org/abs/2606.07271 + Fri, 05 Jun 2026 13:46:37 +0000 - PaperFlow: Profiling, Recommending, and Adapting Across Daily Paper Streams - https://arxiv.org/abs/2606.07454 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07454.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Fuqiang Wang, Song Tan, Zheng Guo, Jiaohao Fu, Xinglong Xu, Bihui Yu, Jie Dong, Zheng Sun, Siyuan Li, Jingxuan Wei, Cheng Tan</p><p><b>Upvotes:</b> 8</p><p><b>Summary:</b> Scientific paper recommendation is typically evaluated as static ranking over a fixed candidate set, yet real scientific reading unfolds as a daily, longitudinal process in which interests shift and feedback accumulates. We introduce PaperFlow, a framework that organizes it into three coupled stages: Profiling, which constructs and maintains a structured, inspectable scholarly profile from heterogeneous cold-start evidence; Recommending, which ranks each date-specific paper stream through multi-signal aggregation under a fixed display budget; and Adapting, which updates user state from semantically distinct feedback signals and models interest drift across days. We further define a longitudinal user-day benchmark that fixes users, dates, candidate pools, visible inputs, and hidden simulated relevance labels under a shared temporal information boundary. The benchmark contains 24 simulated research users, 50 daily paper streams, 1,200 user-day episodes, 20,727 unique papers, and 497,448 episode-paper records. We additionally specify a blind human-evaluation protocol to validate alignment between automatic metrics and expert judgments. Experiments against five scientific recommendation baselines show that PaperFlow achieves the strongest oracle-based ranking, the highest behavioral alignment with simulated reading selections, and the best blind human-evaluation score.</p> - https://arxiv.org/abs/2606.07454 - Fri, 05 Jun 2026 00:00:00 +0000 + Skill-3D: Evolving Scene-Aware Skills for Agentic 3D Spatial Reasoning + https://arxiv.org/abs/2606.07436 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07436.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoyuan Li, Zhengdong Hu, Jun Wang, Hehe Fan, Yi Yang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> This paper explores agentic 3D spatial understanding, i.e., MLLM agents performing 3D reasoning through tool use. Existing methods often misuse tools and exhibit biased tool preferences under 3D scenarios, leaving the agentic paradigm with only marginal gains over non-agentic strategies. We reveal that 3D spatial reasoning tasks are heterogeneous across scenes, while these agents apply a uniform tool-use strategy to all scenes rather than selecting tools according to the specific scene and task. To address this, we propose Skill-3D, a framework that learns self-evolving scene-aware skills. Specifically, Skill-3D identifies the task scene and records the agent's tool-use trajectory into a Scene Memory, where successful trajectories from similar scenes are aggregated and distilled into a reusable scene-aware skill, with failed ones attached to the skill as lessons. During training, once a similar scene recurs, the corresponding skill is injected to guide the agent, producing new trajectories whose successes and failures further refine the skill, forming a loop in which the memory and the skill library co-evolve. Experiments show that Skill-3D substantially improves tool utilization in 3D spatial reasoning (from 39% to 78% on VSI-Bench), driving the agent toward correct and sufficient tool use. For instance, it improves Gemini-3-Flash by 67% on MMSI-Bench. Furthermore, we conduct agentic post-training over skill-guided trajectories, which boosts Qwen3-VL-8B by 43% on VSI-Bench.</p> + https://arxiv.org/abs/2606.07436 + Fri, 05 Jun 2026 16:33:44 +0000 - dots.tts Technical Report - https://arxiv.org/abs/2606.07080 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07080.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Shi Lian, Changtao Li, Bohan Li, Hankun Wang, Da Zheng, Junfeng Tian, Yufeng Ma, Colin Zhang, Kai Yu</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> We present dots.tts, a 2B-parameter continuous autoregressive text-to-speech (TTS) foundation model that models speech in a continuous latent space. Compared with existing continuous autoregressive models, our key innovations are threefold. First, we train an AudioVAE with multiple objectives to build a semantically structured and prediction-friendly continuous speech space. Second, we use full-history conditioning in the flow-matching head to preserve long-range consistency and reduce drift during generation. Third, we apply reward-free self-corrective post-training to the flow-matching head to further improve robustness and acoustic quality. After being trained on a large-scale multilingual corpus, dots.tts achieves the best average performance on Seed-TTS-Eval, with WERs of 0.94%/1.30%/6.60% and SIM scores of 81.0/77.1/79.5 on the zh/en/zh-hard test sets, respectively. Across other benchmarks, dots.tts also consistently demonstrates open-source state-of-the-art performance, exhibiting strong generation stability, voice cloning ability, and emotional expressiveness. For efficient inference, we further apply CFG-aware MeanFlow distillation, enabling low-latency speech generation with first-packet latencies of 85/54 ms in output streaming and dual-streaming modes, respectively. To facilitate reproducible research and practical deployment, we release the training and inference code, together with the pretrained, post-trained, and MeanFlow-distilled checkpoints, under the Apache 2.0 license.</p> - https://arxiv.org/abs/2606.07080 - Fri, 05 Jun 2026 00:00:00 +0000 + Set-Based Transformer for Atmospheric Compensation in Standoff LWIR Hyperspectral Imaging + https://arxiv.org/abs/2606.08324 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08324.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Fabian Perez, Nicolas Quintero, Jeferson Acevedo, Hoover Rueda-Chacon</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Passive long-wave infrared (LWIR) hyperspectral imaging under a standoff geometry depends on atmospheric absorption and emission, as well as reflected radiance, thus making atmospheric compensation essential to get knowledge of a target of interest. Despite its importance, this compensation has been largely overlooked due to its practical and modeling difficulty. In this paper, we present a lightweight set-based deep learning framework that takes multiple radiance measurements, collected at different standoff ranges, as input and jointly estimates transmittance, atmospheric path radiance, and a shared downwelling spectrum. We analyze the learned representation with a sparse autoencoder and observe that several latent features do activate on geographically coherent subsets of the test data despite the absence of location supervision. Experiments on a MODTRAN generated standoff LWIR dataset demonstrate low spectral distortion across all estimated products. The dataset and code is publicly available at: https://factral.co/SAE-LWIR/</p> + https://arxiv.org/abs/2606.08324 + Sat, 06 Jun 2026 00:00:00 +0000 - Watch, Remember, Reason: Human-View Video Understanding with MLLMs - https://arxiv.org/abs/2606.07433 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07433.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiahao Meng, Yue Tan, Qi Xu, Kuan Gao, Weisong Liu, Yanwei Li, Jason Li, Lingdong Kong, Haochen Wang, Qianyu Zhou, Jiangning Zhang, Guangliang Cheng, Yunhai Tong, Lu Qi, Minghsuan Yang</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Video understanding is being rapidly transformed by multimodal large language models (MLLMs), as research moves from short clips to long, multimodal, and knowledge-intensive video scenarios. These scenarios require models to handle sparse evidence, long-range dependencies, multimodal alignment, and reliable inference under limited computational budgets. This work presents a human-view perspective on LLM-based video understanding, organized around three functional abilities: watching, remembering, and reasoning. Rather than treating video tasks as isolated benchmarks, this view provides a unified structure for analyzing how video MLLMs acquire evidence, preserve context, and produce grounded outputs. We introduce a formulation that characterizes video understanding systems by their perceptual representations, memory states, reasoning traces, and final predictions. Based on this formulation, we identify challenges in spatio-temporal perception, efficient long-video processing, memory modeling, streaming understanding, and faithful reasoning. Representative methods are organized by their roles in video MLLM systems. Watching covers fine-grained, comprehensive, audio-visual, and efficient perception. Remembering includes offline and streaming memory, while reasoning covers text-only reasoning and thinking with videos. We further examine application domains such as egocentric, sports, instructional, medical, and narrative videos, and cover training datasets and evaluation benchmarks across task types, supervision formats, modalities, and capability dimensions. Finally, we outline open problems and future directions for scalable, memory-aware, and evidence-grounded video intelligence. Related works will be continuously traced at https://github.com/marinero4972/Awesome-HumanView-VideoUnderstanding.</p> - https://arxiv.org/abs/2606.07433 - Fri, 05 Jun 2026 00:00:00 +0000 + Chiaroscuro Attention: Spending Compute in the Dark + https://arxiv.org/abs/2606.08327 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08327.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Prateek Kumar Sikdar</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Standard transformers apply self-attention uniformly at every layer and token, regardless of whether the input requires dynamic cross-token interaction. We propose CHIAR-Former (Chiaroscuro Attention), a 4-layer hybrid transformer that routes each token to one of three operators - DCT spectral mixing, RBF kernel mixing, or full self-attention - based on per-token spectral entropy, a theoretically justified complexity signal. Through systematic ablation on WikiText-103, we discover routing collapse: the router consistently rejects RBF in favour of DCT and attention, revealing that spectral mixing and dynamic attention are complementary and sufficient. A purpose-designed DCT+Attention-only variant achieves Val PPL 36.54 on WikiText-103 - a 45% improvement over a full-attention baseline (PPL 66.62) at 62.5% fewer attention FLOPs. We extend evaluation to WikiText-2, IMDB sentiment classification, and synthetic ListOps operations, establishing a clear operating regime: CHIAR-Former excels on large-scale naturalistic text where token diversity supports spectral specialisation, while full attention retains an edge on small datasets and synthetic pattern-matching tasks. These findings - both the wins and the losses - together define when and why spectral routing earns its keep.</p> + https://arxiv.org/abs/2606.08327 + Sat, 06 Jun 2026 00:00:00 +0000 - LIMMT: Less is More for Motion Tracking - https://arxiv.org/abs/2606.06953 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.06953.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yu Guan, Zekun Qi, Chenghuai Lin, Xuchuan Chen, Dairu Liu, Wenyao Zhang, Jilong Wang, Xinqiang Yu, He Wang, Li Yi</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> We argue that high-quality motion data can steer tracking policies toward better optimization trajectories early in training. In this work, we introduce LIMMT (Less Is More for Motion Tracking). To our knowledge, this is the first data-centric study for physics-based humanoid motion tracking. We go beyond simply removing low-quality and erroneous clips, but define motion data quality through three dimensions: physics feasibility, diversity, and complexity. We show that even training with under 3% of AMASS yields better tracking performance than training with the full dataset. We further conduct data cleaning on the estimated web-sourced mocap data. Extensive experiments and analyses validate the effectiveness of our framework.</p> - https://arxiv.org/abs/2606.06953 - Fri, 05 Jun 2026 00:00:00 +0000 + EmpiriGraph-Psy: A Dataset and LLM Pipeline for Extracting Empirical Relation Graphs from Psychology Abstracts + https://arxiv.org/abs/2606.08362 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08362.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Danqin Zhao, Yicun Liu, Xingwei Tan, Thomas T. Hills</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Existing scientific relation extraction benchmarks mainly target domains such as computer science, where entities are tasks, methods, datasets, materials, or metrics. This leaves a gap in variable-oriented empirical fields such as psychology, where findings are expressed as relations among constructs, measurements, interventions, and outcomes. We introduce variable-centered empirical graph extraction, the task of mapping scientific abstracts to typed graphs whose nodes are normalized variables and whose edges represent empirical and hierarchical relations. To support this task, we construct EmpiriGraph-Psy, a benchmark of 210 psychology abstracts annotated by domain-trained annotators with normalized variables, concept hierarchies, empirical relation types, and validation states. We evaluate frontier and open-weight LLMs using both direct extraction and a staged graph-construction pipeline that separates variable extraction, normalization, hierarchy construction, evidence selection, relation extraction, and edge validation. The staged pipeline substantially outperforms direct extraction, with the best configuration achieving a macro-F1 of 0.74. Error analysis shows that moderation relations and concept hierarchies remain the most challenging cases, highlighting the difficulty of extracting higher-order empirical claims and implicit abstraction structure from scientific abstracts.</p> + https://arxiv.org/abs/2606.08362 + Sat, 06 Jun 2026 00:00:00 +0000 - UniSHARP: Universal Sharp Monocular View Synthesis - https://arxiv.org/abs/2606.07514 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07514.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Meixi Song, Dizhe Zhang, Hao Ren, Ruiyang Zhang, Bo Du, Ming-Hsuan Yang, Lu Qi</p><p><b>Upvotes:</b> 13</p><p><b>Summary:</b> In this work, we focus on extending SHARP, the popular photorealistic view synthesis method, for universal monocular rendering across a continuum of camera systems, from conventional perspective cameras to wide-field-of-view, fisheye and omnidirectional panoramic settings. To overcome the pinhole-specific assumptions of SHARP, our key idea is to align various images in a unified omnidirectional latent space. Thus, we propose UniSHARP, which performs implicit alignment in both feature and Gaussian spaces. Specifically, Gaussian primitives are arranged along rays and radial distances in a ray-based universal representation, while 2D semantic and 3D spatial features extracted from UniK3D-inspired encoders are jointly decoded to generate the complete Gaussian cloud. To comprehensively evaluate our method, we construct a benchmark covering diverse imaging systems across various scenes. The benchmark is further stratified by field of view (FoV) to enable fine-grained assessment of the universal monocular rendering task. Extensive experiments on the proposed benchmark demonstrate the effectiveness of UniSHARP, outperforming alternative methods by a large margin. The project page can be found at: https://insta360-research-team.github.io/Unisharp-website/</p> - https://arxiv.org/abs/2606.07514 - Fri, 05 Jun 2026 00:00:00 +0000 + Light-WAM: Efficient World Action Models with State-Fusion Action Decoding + https://arxiv.org/abs/2606.08242 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08242.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ziang Li, Dongzhou Cheng, Yibin Wang, Shiyue Wang, Xiaoyang Xu, Lingxuan Weng, Juan Wang, Jiaqi Wang</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> World Action Models (WAMs) extend robot policy learning by incorporating future prediction as an additional training objective, encouraging the policy to encode task-relevant temporal structure in its representations. Current WAMs often rely on large-scale generative architectures that incur high training costs and inference latency, making them difficult to deploy as efficient closed-loop policies. We propose Light-WAM, a lightweight World Action Model for efficient robot manipulation. Specifically, it is built with a compact video backbone and performs future-video supervision in a downsampled latent space, reducing the cost of video co-training while retaining its benefits for representation learning. For action prediction, Light-WAM introduces the StateFusionActionExpert, which reads adapted states from multiple backbone layers, fuses them through learned-query pooling, and directly predicts action chunks in a single forward pass. This design provides an efficient interface between video backbone representations and robot actions, avoiding the need for heavy generative action experts. Experiments demonstrate that Light-WAM maintains strong performance on LIBERO and achieves usable multi-task performance on RoboTwin 2.0, while using only 0.44B trainable parameters. It also achieves 72.03ms inference latency with 4.1GiB peak GPU memory and improved training throughput.</p> + https://arxiv.org/abs/2606.08242 + Sat, 06 Jun 2026 00:00:00 +0000 - AnchorWorld: Embodied Egocentric World Simulation with View-based Evolution Customization - https://arxiv.org/abs/2606.07326 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07326.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yu Li, Menghan Xia, Gongye Liu, Xintao Wang, Conglang Zhang, Lei Ke, Yuxuan Lin, Ruihang Chu, Pengfei Wan, Kun Gai, Yujiu Yang</p><p><b>Upvotes:</b> 24</p><p><b>Summary:</b> Despite being a pivotal frontier, interactive world modeling remains underexplored in terms of the versatile controllability required by practical scenarios. To bridge this gap, we present AnchorWorld, a framework that advances egocentric simulation through enhanced interaction integrity and a flexible mechanism for world customization. First, we utilize 3D human motion as the primary interaction modality. To complement the out-of-view or truncated body parts in egocentric views, we introduce an auxiliary training supervision that incorporates exogenous viewpoints decoupled from the agent's first-person sensorium. It allows the model to observe the agent's full-body positioning relative to the environment, facilitating a more robust spatial grounding of human-world interactions. Furthermore, we propose a simple yet effective mechanism for customizing self-evolving worlds. This is achieved by defining anchor views within a unified world coordinate system, coupled with textual descriptions dictating the dynamic evolution of local scenes. Experimental results show that AnchorWorld significantly outperforms state-of-the-art baselines, while ablation studies validate the effectiveness of our key designs. Notably, our customization scheme exhibits promising spatio-temporal geometric consistency and adheres strictly to the prescribed evolutionary dynamics.</p> - https://arxiv.org/abs/2606.07326 - Fri, 05 Jun 2026 00:00:00 +0000 + Bayesian-Agent: Posterior-Guided Skill Evolution for LLM Agent Harnesses + https://arxiv.org/abs/2606.08348 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08348.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Xiaojun Wu, Cehao Yang, Honghao Liu, Xueyuan Lin, Wenjie Zhang, Zhichao Shi, Xuhui Jiang, Chengjin Xu, Jia Li, Jian Guo</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> LLM agents increasingly rely on external inference conditions: prompts, tools, memory, SOPs, skills, and harness feedback. These assets can improve task execution without changing model weights, but they are often revised by heuristic reflection or by reusing observed successes and failures as if counts alone were reliable belief. We introduce Bayesian-Agent, a native and cross-harness framework that treats reusable skills and SOPs as hypotheses about whether a frozen model will succeed under a particular prompt, context, and harness environment. Bayesian-Agent records verified trajectory evidence, maintains a feature-conditioned categorical posterior over each skill, and maps posterior state into inspectable actions such as patch, split, compress, retire, and explore. Model-facing prompts receive executable guardrails and failure-mode patches, while posterior summaries remain available for audit. With deepseek-v4-flash, incremental repair improves SOP-Bench from 80\% to 95\%, Lifelong AgentBench from 90\% to 100\%, and RealFin-Bench from 45\% to 65\%. We further evaluate Bayesian-Agent's native backend and optional GenericAgent, mini-swe-agent, and Claude Code backends. The results include positive, negative, saturated, and case-study settings, suggesting that agent skill evolution is best viewed as posterior-guided harness optimization rather than uncalibrated prompt accumulation. The source code is available at https://github.com/DataArcTech/Bayesian-Agent.</p> + https://arxiv.org/abs/2606.08348 + Sat, 06 Jun 2026 00:00:00 +0000 - MMAE: A Massive Multitask Audio Editing Benchmark - https://arxiv.org/abs/2606.07229 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07229.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ziyang Ma, Ruiqi Yan, Ruiyang Xu, Jie Fang, Zhikang Niu, Yi-Wen Chao, Wenming Tu, Tianrui Wang, Auden, Qi Chen, Wenxi Chen, Jiaying Chi, Yanru Huo, Zixuan Jiang, Xiquan Li, Yalin Li, Junxi Liu, Minghao Liu, Binghao Qiang, Yijia Shan, Zheshu Song, Tian Tan, Zixiang Wang, Zeyu Xie, Zhifei Xie, Xiaoyu Xing, Qixiang Xu, Chen Yang, Guanrou Yang, Shan Yang, Yifan Yang, Steve Yves, Haotian Zhang, Haina Zhu, Kai Yu, Liefeng Bo, Eng-Siong Chng, Xie Chen</p><p><b>Upvotes:</b> 39</p><p><b>Summary:</b> We introduce MMAE, a Massive Multitask Audio Editing benchmark, serving as the first comprehensive evaluation testbed designed for general-purpose instruction-based audio editing. Spurred by the shift toward intelligent creation, interactive editing has rapidly expanded from visual domains, pioneered by models like Nano-banana 2 for images and Gemini-Omni for video, into audio. However, the current evaluation infrastructure lags severely, remaining highly fragmented and restricted to specific subdomains or basic operations. Unlike existing benchmarks that are limited in scope, MMAE extends to a broad spectrum of real-world scenarios, encompassing 7 distinct audio modalities, including sound, speech, music, and their mixtures. Furthermore, we establish a comprehensive taxonomy spanning 6 levels of task complexity, from basic modifications to multi-hop reasoning and multi-round editing, 2 levels of granularity, and 8 distinct operation types. Meticulously curated through human-agent collaboration, MMAE comprises 2,000 high-fidelity samples paired with a pioneering rubric-based evaluation framework. By decomposing free-form tasks into 17,741 verifiable criteria, this robust rubric-based paradigm enables a precise, multi-dimensional assessment of both instruction following and context consistency. Our extensive evaluation of leading models reveals that current systems remain far from achieving reliable edits. Strikingly, the Exact Match Rate (EMR) consistently falls below 5% and plummets to an absolute 0% in complex, mixed-modality tasks, exposing critical bottlenecks in precise execution and structural robustness. We hope MMAE will serve as a catalyst for future advances in the intelligent creation community, providing a clear diagnostic roadmap and establishing a standardized, long-lasting evaluation paradigm for next-generation audio editing systems.</p> - https://arxiv.org/abs/2606.07229 - Fri, 05 Jun 2026 00:00:00 +0000 + Phase Marginalization for Patch-Grid Instability in Vision Transformers + https://arxiv.org/abs/2606.08132 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08132.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Oğuzhan Ercan</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Vision Transformers operate on fixed patch grids, which can introduce phase-dependent instability for dense prediction: changing the patch partition can change the token evidence available to a pixel, especially near boundaries. We formalize patch-grid phase as a nuisance variable and propose Phase Marginalization, a post-hoc marginalization method that evaluates structured patch-grid phases, inverse-aligns dense outputs, and aggregates them in the original image coordinate system. The central variant, Uniform Phase Marginalization with K = 4, is training-free and improves over the canonical K = 1 baseline across measured segmentation, depth, and local matching settings. In a controlled Cityscapes experiment, Uniform Phase Marginalization provides a modest compute-matched advantage over generic shift-based four-forward test-time augmentation (TTA) (+0.31 mean Intersection-over-Union over the strongest tested generic row). A scaling study further shows that K = 4 is a practical cost-accuracy trade-off: K = 8 is essentially unchanged and K = 16 adds little accuracy at much higher latency. These results position patch-grid phase as a measurable nuisance variable and Phase Marginalization as a simple diagnostic and post-hoc marginalization baseline for dense ViT prediction.</p> + https://arxiv.org/abs/2606.08132 + Sat, 06 Jun 2026 12:15:56 +0000 - Your UnEmbedding Matrix is Secretly a Feature Lens for Text Embeddings - https://arxiv.org/abs/2606.07502 - <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.07502.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Songhao Wu, Zhongxin Chen, Yuxuan Liu, Heng Cui, Cong Li, Rui Yan</p><p><b>Upvotes:</b> 69</p><p><b>Summary:</b> Large language models exhibit impressive zero-shot capabilities across a wide range of downstream tasks. However, they struggle to function as off-the-shelf embedding models, leading to suboptimal performance on massive text embedding benchmarks. In this paper, we identify a potential cause underlying this deficiency. Our motivation stems from an unexpected observation: text embeddings tend to align with frequent but uninformative tokens when projected onto the vocabulary space. We argue that this excessive expression of high-frequency tokens suppresses the model's ability to capture nuanced semantics. To address this, we introduce EmbedFilter, a simple linear transformation designed to refine text embeddings derived from LLMs directly. Specifically, we uncover that the unembedding matrix within LLMs encodes a latent space that is actively writing these frequent tokens into embedding space. By filtering out this subspace, EmbedFilter suppress the influence of high-frequency tokens, thereby enhancing semantic representations. As a compelling byproduct, this enables an inherent dimensionality reduction, lowering index storage and speedup retrieval while fully preserving the refined embedding quality. Our experiments across multiple LLM backbones demonstrate that LLMs equipped with EmbedFilter achieve superior zero-shot downstream performance even with significantly reduced embedding dimensions. We hope our findings provide deeper insights into the mechanisms of LLM-based representations and inspire more principled designs to improve text embeddings training. Our code is available at https://github.com/CentreChen/EmbFilter.</p> - https://arxiv.org/abs/2606.07502 - Fri, 05 Jun 2026 00:00:00 +0000 + PIPE-Cypher: Automatic Enterprise Benchmark Generation for Text-to-Cypher Systems + https://arxiv.org/abs/2606.08481 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08481.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Suraj Ranganath, Anish Raghavendra</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Enterprise property graphs vary widely in schema structure, internal terminology, domain assumptions, governance constraints, and user interaction patterns. A deployment-relevant Text2Cypher benchmark therefore reflects the questions users and agents actually ask of that graph. Creating such a benchmark is difficult because schemas and values are unique, and graph structure changes over time. Each NL-query pair must also be executable, use real graph entities, preserve diversity, and remain balanced across query types and difficulty levels. We present PIPE-Cypher, a local benchmark-generation pipeline that turns a live property graph and optional seed queries from customer questions, analyst logs, or agent tool calls into balanced NL-to-Cypher benchmarks. PIPE-Cypher combines schema profiling, reverse-query grounding, constrained generation, deterministic Cypher governance, execution validation, redaction, diversity controls, and a calibrated local LLM judge. Using local Qwen3.5-9B generation and judging, PIPE-Cypher exports 3,000 accepted FinBench/SNB examples, completes three audited ablation suites, calibrates judge behavior with human labels, and evaluates 11 local downstream models. The resulting benchmark is deliberately discriminative: zero-shot transfer is weak, while a few-shot control shows that schema-specific example banks can help compatible model families. Together, PIPE-Cypher makes Text2Cypher benchmarking a repeatable process that evolves with the graph, its users, and its target workloads.</p> + https://arxiv.org/abs/2606.08481 + Sun, 07 Jun 2026 00:00:00 +0000 + + + OASIS: From Simulation Data Collection to Real-World Humanoid Loco-Manipulation + https://arxiv.org/abs/2606.08548 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08548.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Zehao Yu, Jiakun Zheng, Weiji Xie, Jiyuan Shi, Chenyun Zhang, Chenjia Bai, Xuelong Li</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Recent progress in robot manipulation has been largely driven by learning from large-scale demonstrations. For humanoid robot loco-manipulation tasks, however, existing data sources force an unsatisfying tradeoff between trajectory quality and scalability. Real-world teleoperation provides the highest-quality trajectories but requires dedicated physical space and time-consuming scene resets. Simulation offers an alternative way out of this dilemma: it can produce clean, embodiment-aligned data at scale without any physical hardware. In this paper, we propose OASIS, a simulation-data-driven framework for humanoid loco-manipulation. OASIS automatically reconstructs realistic object assets from real-world images using a 3D generative model. Based on these assets, trajectories are first collected through teleoperation in simulation, and then augmented under diverse domain randomizations in a post-processing stage. With the resulting simulation data, we further design a hierarchical visuomotor policy for humanoid loco-manipulation. Extensive experiments on the real humanoid robot show that, under zero-shot deployment, the policy trained on our simulation data achieves higher success rates on most tasks than that trained on real-robot teleoperation data, owing largely to the broad lighting and environmental variations covered by our simulation rendering, which real-robot data fails to capture. The project page is available at https://oasis-humanoid.github.io/.</p> + https://arxiv.org/abs/2606.08548 + Sun, 07 Jun 2026 00:00:00 +0000 + + + Trajectory-Refined Distillation + https://arxiv.org/abs/2606.08432 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08432.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Li Jiang, Haoran Xu, Yichuan Ding, Amy Zhang</p><p><b>Upvotes:</b> 4</p><p><b>Summary:</b> On-policy distillation (OPD) has become a central post-training tool for large language models (LLMs), providing dense per-token teacher supervision along the student's own rollouts. In this work, we identify a common structural cause underlying OPD, which we call prefix failure. Under prefix failure, dense per-token supervision induces a bimodal teacher mixture and fragmented gradients that token-level loss truncation or reweighting fail to address. This observation motivates us to move beyond token-level loss interventions toward trajectory-level output corrections. We thus propose Trajectory-Refined Distillation (TRD), a trajectory-level correction method that revises the student's rollout under the teacher guidance while within on-policy support. By correcting problematic prefixes before distillation, TRD mitigates prefix failure at its source. Moreover, TRD improves the exploration by exposing the student to alternative valid derivations under teacher guidance, even when the original rolls are already correct. TRD can also be applied to on-policy self-distillation (OPSD), a parameter-sharing variant that uses the student model conditioned on privileged informations as the teacher. Across a wide range of benchmarks and base models at multiple scales, TRD consistently outperforms prior baselines, improving single-attempt accuracy and broadening reasoning coverage. Code is available at https://github.com/louieworth/trd</p> + https://arxiv.org/abs/2606.08432 + Sun, 07 Jun 2026 00:00:00 +0000 + + + OmniCap-IF: Benchmarking and Improving Instruction Following Abilities for Omni-Video Captioning + https://arxiv.org/abs/2606.08572 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08572.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiahao Wang, An Ping, Yanghai Wang, Yuanxing Zhang, Shihao Li, Hanyan Bian, Yichi Ren, Yize Zhang, Han Wang, Haowen Chen, Junze Li, Jiaqi Wang, Yiyang Hu, Zhuze Xu, Zijie Zhang, Jiaheng Liu</p><p><b>Upvotes:</b> 9</p><p><b>Summary:</b> While Omni-modal Large Language Models (OLLMs) have demonstrated impressive capabilities in jointly processing audio and visual streams, their ability to strictly adhere to complex, multi-faceted user instructions remains largely unexplored. Existing benchmarks primarily focus on holistic video understanding or text-only instruction following, failing to capture the intricate interplay between modalities and user constraints. To bridge this gap, we introduce OmniCap-IF, the first comprehensive benchmark specifically designed to evaluate instruction-following capabilities in omni-modal captioning. OmniCap-IF incorporates a systematic framework that assesses captions on two dimensions: format correctness and content correctness. Our benchmark encompasses 50 distinct constraint types across pure visual, pure audio, and audio-visual modalities, while integrating Temporal Grounding to assess spatio-temporal precision. Extensive evaluations of prominent models on 1,920 high-quality samples reveal significant performance disparities. Furthermore, our analysis uncovers a critical "format-content tradeoff", demonstrating that increasing formatting complexity directly degrades models' omni-modal reasoning abilities. Finally, to advance the field, we curate a 54K instruction-tuning dataset, OmniCap-IF-54K and present OmniCaptioner-IF, which achieves notable improvements in both complex instruction adherence and general omni-modal captioning performance.</p> + https://arxiv.org/abs/2606.08572 + Sun, 07 Jun 2026 00:00:00 +0000 + + + CoVEBench: Can Video Editing Models Handle Complex Instructions? + https://arxiv.org/abs/2606.08415 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08415.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiangtao Wu, Jiaming Wang, Yiwen He, Yuanxing Zhang, Shihao Li, Dunyuan Liu, Xuedong Zhao, Jialu Chen, Zekun Moore Wang, Jiaheng Liu</p><p><b>Upvotes:</b> 44</p><p><b>Summary:</b> While recent text-guided video editing models excel at elementary tasks (e.g., style transfer, object insertion), real-world user requests are highly compositional. A single prompt often demands multiple coupled edits, such as modifying subjects, actions, and camera views, while strictly preserving unrelated spatiotemporal content. Existing benchmarks, heavily constrained by isolated edits and coarse global metrics, fail to diagnose how models handle such complex workflows. To address this gap, we introduce CoVEBench, a compositional video editing benchmark comprising 416 curated source videos, 626 multi-point editing instructions, and 9,990 fine-grained checklist items. Covering diverse editing dimensions, CoVEBench evaluates models via MLLM-judged instruction compliance and video fidelity, alongside automated metrics for video quality. Extensive experiments reveal that compositional editing remains a profound challenge: current models frequently omit edits, violate preservation constraints, or introduce artifacts when handling multiple operations simultaneously. CoVEBench provides a challenging, diagnostic testbed to advance video editing toward realistic user workflows.</p> + https://arxiv.org/abs/2606.08415 + Sun, 07 Jun 2026 00:00:00 +0000 + + + Precision Is Not Faithfulness: Coverage-Aware Evaluation of Grounded Generation with a Complete Oracle + https://arxiv.org/abs/2606.09376 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09376.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Juan S. Santillana</p><p><b>Upvotes:</b> 0</p><p><b>Summary:</b> Reference-free faithfulness metrics verify each atomic claim a model makes against ground truth, and are increasingly used to evaluate grounded generation. We show they share a blind spot: they measure only precision -- are the stated claims supported? -- and therefore reward abstention, since a model can score near-perfect faithfulness by saying almost nothing. We make this measurable using Formula 1 telemetry, a domain where strategic ground truth is derived deterministically and, crucially, completely: for each decision we know the full set of facts that mattered. This completeness -- absent in open-domain faithfulness benchmarks -- lets us measure recall (coverage of the relevant facts) exactly, alongside precision. On a multilingual (EN/ES/PT) benchmark of 7,253 decision instances spanning 150 races, the most precise frontier model covers under half of the relevant facts and ranks last by F1, so requiring coverage reorders the systems; the same effect reappears in a second complete-oracle domain (NOAA weather forecasts). A prompt ablation shows the low coverage is not an under-prompting artifact: explicitly asking models to be thorough does not close the gap. We pair faithfulness with coverage into a single score, validate the metric (controlled perturbation; agreement across a model-free regex extractor and a cross-family LLM extractor, system-level Spearman 1.0), and give a verifier-guided generation method that improves precision and recall without references. We release the benchmark, structured annotations, metric, baselines, and an interactive demo.</p> + https://arxiv.org/abs/2606.09376 + Mon, 08 Jun 2026 00:00:00 +0000 + + + PBSD: Privileged Bayesian Self-Distillation for Long-Horizon Credit Assignment + https://arxiv.org/abs/2606.09348 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09348.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yang Tian, Rui Wang, Xumeng Wen, Junjie Li, Shizhao Sun, Lei Song, Jiang Bian, Bo Zhao</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Long-horizon agentic tasks pose a fundamental credit assignment challenge for outcome-base reinforcement learning: trajectory-level rewards verify final correctness but provide limited guidance on which intermediate reasoning steps or tool interactions contribute to the outcome. The difficulty is especially pronounced in multi-turn search agents, where successful trajectories may contain misleading actions and failed trajectories may contain valuable evidence-gathering steps. We propose PBSD (Privileged Bayesian Self-Distillation), a Bayes-calibrated self-distillation method for fine-grained credit assignment under sparse final rewards. PBSD measures trajectory quality through the posterior-to-prior probability ratio of the verified answer and applies Bayes' rule to convert this hard-to-estimate answer-side ratio into a tractable likelihood ratio between a standard student model and a privileged answer-conditioned teacher model. Autoregressive decomposition of this Bayesian evidence score yields turn-level signals that identify whether each intermediate turn supports or undermines the verified outcome. Consequently, PBSD provides a principled and elegant reweighting scheme that transforms sparse outcome supervision into Bayes-calibrated turn-level credit signals, while remaining fully compatible with standard policy optimization. Experiments demonstrate that PBSD consistently enhances performance across both in-domain and out-of-domain settings, and effectively transfers knowledge from short-context training to long-context inference, suggesting that its fine-grained credit assignment mechanism facilitates more effective policy learning and yields improved generalization.</p> + https://arxiv.org/abs/2606.09348 + Mon, 08 Jun 2026 00:00:00 +0000 + + + Evaluation Cards: An Interpretive Layer for AI Evaluation Reporting + https://arxiv.org/abs/2606.09809 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09809.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Avijit Ghosh, Anka Reuel, Jenny Chim, Wm. Matthew Kennedy, Srishti Yadav, Jennifer Mickel, Yanan Long, Andrew Tran, Anastassia Kornilova, Damian Stachura, Kevin Klyman, Felix Friedrich, Jeba Sania, Max Lamparth, Jan Batzner, Anoop Mishra, Eliya Habba, Yixiong Hao, Nathan Heath, Shalaleh Rismani, Usman Gohar, Andrea Loehr, David Manheim, Ruchira Dhar, Sree Harsha Nelaturu, Aarush Sinha, Leshem Choshen, Drishti Sharma, Ishan Khire, Amit Saha, Subramanyam Sahoo, Michael Hardy, Michael Alexander Riegler, Kabir Manghnani, Michelle Lin, Yanan Jiang, Yilin Huang, Asaf Yehudai, Jessica Ji, Aris Hofmann, Mubashara Akhtar, Nuno Moniz, Yacine Jernite, Stella Biderman, Zeerak Talat, Sanmi Koyejo, Mykel Kochenderfer, Irene Solaiman</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> AI evaluation results are produced at scale but reported inconsistently across leaderboards, model cards, benchmark papers, and company blogs. The cost is interpretive: readers cannot reliably compare results across sources, identify what a report omits, or trace an aggregate claim to its underlying evidence. Recent efforts address isolated components but leave three gaps: they cover only narrow slices of the evaluation lifecycle and do not compose into a single interpretable record; they specify static representations that do not differentiate the questions different stakeholders bring to the same evidence; and they remain proposals on paper, lacking the extraction infrastructure required for adoption at scale. We present , an operational reporting layer that composes benchmark metadata, evaluation run data, and model metadata into a unified record. We (1) derive a reporting schema from a structured review of 52 papers and 10 stakeholder interviews, (2) implement four interpretive signals (reproducibility, documentation completeness, provenance and risk, and score comparability), rendered through reader modes calibrated to research and non-research audiences, and (3) deploy a monitoring tool that applies across 5,816 models, 635 benchmarks, and 101,843 results, surfacing systematic gaps in current reporting practice.</p> + https://arxiv.org/abs/2606.09809 + Mon, 08 Jun 2026 00:00:00 +0000 + + + SwiftVR: Real-Time One-Step Generative Video Restoration + https://arxiv.org/abs/2606.09516 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09516.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jiaqi Yan, Xiangyu Chen, Xinlin Zhong, Haibin Huang, Chi Zhang, Jie Liu, Jiantao Zhou, Xuelong Li</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> Real-time video restoration (VR) for live streams requires high-resolution outputs under strict per-frame latency constraints. Existing one-step diffusion-based VR models remain difficult to deploy on consumer-grade GPUs due to two main bottlenecks: quadratic spatial attention at high resolutions and the latency-memory overhead of large video autoencoders. We present SwiftVR, a streaming one-step generative VR framework that reduces both bottlenecks under a causal chunk-wise protocol. For attention, mask-free shifted-window self-attention gathers each spatial window into a dense tensor via deterministic indexing, keeping all attention calls on the dense scaled dot-product attention path without masks, cyclic shifts, padding, or hardware-specific sparse kernels. Because SwiftVR uses only standard dense SDPA calls, the trained model transfers to consumer GPUs without retraining or custom kernels. For autoencoding, a lightweight Restoration-aware Autoencoder enables fast chunk-wise decoding while preserving reconstruction quality. On a single H100, SwiftVR sustains 31~FPS at 2560x1440 and 14~FPS at 3840x2160, whereas all compared diffusion-based VR baselines exceed the memory limit at 4K. On a consumer RTX~5090, SwiftVR reaches 26~FPS at 1920x1080. To our knowledge, SwiftVR is the first generative VR model to achieve real-time 1080p streaming on a consumer-grade GPU, while attaining strong no-reference perceptual quality with lower inference cost. Project is available at https://h-oliday.github.io/SwiftVR.</p> + https://arxiv.org/abs/2606.09516 + Mon, 08 Jun 2026 00:00:00 +0000 + + + AHA-WAM:Asynchronous Horizon-Adaptive World-Action Modeling with Observation-Guided Context Routing + https://arxiv.org/abs/2606.09811 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09811.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Jisong Cai, Long Ling, Shiwei Chu, Zhongshan Liu, Jiayue Kang, Zhixuan Liang, Wenjie Xu, Yinan Mao, Weinan Zhang, Xiaokang Yang, Ru Ying, Ran Zheng, Yao Mu</p><p><b>Upvotes:</b> 12</p><p><b>Summary:</b> World-action models have emerged as a promising paradigm for robot manipulation, jointly modeling visual scene dynamics and actions to inject physical priors into policy learning. However, existing world-action models couple world prediction and action execution at the same temporal resolution, forcing the world branch to model near-term frame variations that are redundant and weakly informative. We posit that strictly binding world prediction and action execution to the same temporal rhythm may underutilize the potential of the video branch for embodied control. Therefore, we propose AHA-WAM, an Asynchronous Horizon-Adaptive World-Action Model built on a dual Diffusion Transformer (DiT) architecture that reorganizes world-action modeling around this temporal asymmetry. AHA-WAM instantiates the video DiT as a low-frequency world planner that maintains rolling key-value memory over past observations and exposes reusable layerwise latent context encoding long-horizon scene evolution, while a high-frequency action DiT executes short action chunks in closed loop by querying this context through layerwise joint attention. To support asynchronous execution, we introduce horizon-adaptive offset training and Observation-Guided Video-Context Routing (OVCR), which together let the action expert exploit long-horizon world context while remaining responsive to real-time execution state without rerunning the video DiT. Experiments on RoboTwin and real-world manipulation tasks show that AHA-WAM achieves state-of-the-art performance without any robot-data pretraining, attaining 92.80% average success on RoboTwin and 78.3% success across 4 real-world tasks, while reaching 24.17 Hz closed-loop control with a 4.59x speedup over Fast-WAM.</p> + https://arxiv.org/abs/2606.09811 + Mon, 08 Jun 2026 00:00:00 +0000 + + + OmniGameArena: A Unified UE5 Benchmark for VLM Game Agents with Improvement Dynamics + https://arxiv.org/abs/2606.09826 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09826.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Mingxian Lin, Shengju Qian, Yuqi Liu, Yi-Hua Huang, Yiyu Wang, Wei Huang, Yitang Li, Fan Zhang, Zeyu Hu, Lingting Zhu, Xin Wang, Xiaojuan Qi</p><p><b>Upvotes:</b> 16</p><p><b>Summary:</b> Vision-language model (VLM) agents are increasingly deployed in interactive game environments. Yet game benchmarks for VLM agents typically report a single first-attempt score per (agent, game) pair, focus on single-agent Solo play, and lack unified protocols for evaluating heterogeneous agent classes (commercial VLMs, open-weight VLMs, and specialized game policies) on the same footing. We address these gaps with OmniGameArena, a real-time benchmark of twelve newly built Unreal Engine 5 games spanning Solo (7), PvP (3), and Coop (2) with unified action interfaces, and the Improvement Dynamics Curve (IDC), an agentic-reflection harness in which a tool-using reflector LLM autonomously refines a bounded skill prompt across multiple rounds. Beyond cold-start leaderboard scores, IDC exposes two additional observables for each (agent, game) pair: how the score evolves across reflection rounds, and how the learned skill behaves on held-out task variants. We report these observables for twelve VLM agents on the cold-start leaderboard and four top agents under IDC.</p> + https://arxiv.org/abs/2606.09826 + Mon, 08 Jun 2026 00:00:00 +0000 + + + End-to-End Context Compression at Scale + https://arxiv.org/abs/2606.09659 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09659.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ang Li, Sean McLeish, Haozhe Chen, Nimit Kalra, Zaiqian Chen, Artem Gazizov, Venkata Anoop Suhas Kumar Morisetty, Bhavya Kailkhura, Harshitha Menon, Zhuang Liu, Brian R. Bartoldson, Tom Goldstein, Sanae Lotfi, Micah Goldblum, Pavel Izmailov</p><p><b>Upvotes:</b> 17</p><p><b>Summary:</b> Long-context language model inference is bottlenecked by memory, as the KV cache grows with context length. Recent techniques to compress the KV cache fall short: they either degrade model quality substantially or require considerable time and compute to compress a single long prompt. Furthermore, many methods require the input to fit within the target model's context window, and are generally incompatible with modern production inference engines. Encoder-decoder compressors, which map a long token sequence to a shorter sequence of latent embeddings consumed by a decoder, are an appealing alternative in principle. However, existing approaches are not competitive with KV cache compression on the accuracy-efficiency frontier. In this work, we revisit encoder-decoder compression and close this gap. We first perform an architecture search, pre-training many variants from scratch to determine how best to design and train encoder-decoder compressors. Guided by our findings, we continually pre-train a family of 0.6B-encoder, 4B-decoder models on over 350B tokens each, at compression ratios of 1:4, 1:8, and 1:16. We introduce Latent Context Language Models (LCLMs), a family of compressors that improve the Pareto frontier across general-task performance, compression speed, and peak memory usage. We demonstrate that LCLMs serve as efficient backbones for long-horizon agents, letting the agent skim through a compressed long context and adaptively expand relevant segments on demand.</p> + https://arxiv.org/abs/2606.09659 + Mon, 08 Jun 2026 00:00:00 +0000 + + + Echo-Memory: A Controlled Study of Memory in Action World Models + https://arxiv.org/abs/2606.09803 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09803.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Wayne King, Zeyue Xue, Yuxuan Bian, Jie Huang, Haoran Li, Yaowei Li, Yaofeng Su, Yuming Li, Haoyu Wang, Shiyi Zhang, Songchun Zhang, Yuwei Niu, Sihan Xu, Junhao Zhuang, Haoyang Huang, Nan Duan</p><p><b>Upvotes:</b> 27</p><p><b>Summary:</b> We present Echo-Memory, a controlled study of memory mechanisms in action-conditioned world models. These models generate multi-segment videos from a first frame, text prompt, and camera-action sequence, but their central failure is often memory rather than local image synthesis: after the camera leaves and returns, the scene or salient object may silently change. Existing memory designs are hard to compare because gains are entangled with backbone, training, retrieval, and evaluation differences. Echo-Memory fixes the action-to-video interface and varies only how history is stored and read by the generator. Under a shared video diffusion backbone, optimizer, camera-action representation, sampler, and evaluation pipeline, we compare raw context, compression-based memory, spatial summaries with different read-out paths, and state-space recurrence. This matched matrix separates four otherwise conflated axes: capacity, compression, read-out, and recurrence. We also evaluate memory through a three-branch protocol: replay quality, in-domain loop revisit, and open-domain return probes. The branches routinely disagree, showing that replay fidelity is not a sufficient proxy for remembering a world. Three findings follow. Raw context is a strong capacity baseline and improves open-domain return far more than it improves replay metrics. Compactness is not a free substitute for capacity: aggressive spatial and hybrid-compression memories lose the salient evidence needed for return. Finally, block-wise state-space recurrence is the strongest open-domain return mechanism in our matrix, showing that the structure of implicit memory matters as much as the decision to use it. These results provide a compact protocol for studying memory in action world models beyond isolated replay metrics.</p> + https://arxiv.org/abs/2606.09803 + Mon, 08 Jun 2026 00:00:00 +0000 + + + SpatialWorld: Benchmarking Interactive Spatial Reasoning of Multimodal Agents in Real-World Tasks + https://arxiv.org/abs/2606.09669 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09669.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Hongcheng Gao, Hailong Qu, Jingyi Tang, Jiahao Wang, Zihao Huang, Hengkang Qiao, Shihong Huang, Junming Yang, Yi Li, Hongyixuan Yuan, Wenjie Li, Bohan Zeng, Wenbo Li, Bo Wang, Jianhui Liu, Olive Huang, Haoyang Huang, Wentao Zhang, Guoqing Huang, Nan Duan, Yinpeng Dong</p><p><b>Upvotes:</b> 37</p><p><b>Summary:</b> Spatial reasoning is a foundational capability for multimodal large language models (MLLMs) to perceive and operate within the physical world. However, existing benchmarks predominantly rely on passive evaluation (e.g., static VQA) or simulator-specific pipelines, failing to assess general interactive spatial understanding. We introduce SpatialWorld, a unified benchmark designed specifically for evaluating the interactive spatial understanding of multimodal agents in complex real-world tasks. Integrating eight heterogeneous simulation backends under a shared, simulator-agnostic protocol, SpatialWorld features 760 human-annotated tasks across diverse domains (e.g., household routines, travel, social collaboration). Agents must solve tasks under vision-only partial observability, actively gathering egocentric visual evidence and expressing decisions via a unified, text-based action interface native to MLLMs. For reliable evaluation, each task includes a human-validated initial state, a reference trajectory, and a terminal-state verifier. Evaluating 15 advanced agents reveals that robust spatial task solving remains challenging: the strongest model, GPT-5, achieves an average task success rate (TSR) of only 17.4%, while the leading open-source model, Qwen-3.5, reaches 14.1%. Further analysis exposes a clear mismatch between task success and execution efficiency, alongside substantial domain-specific performance variations. These bottlenecks in active exploration and long-horizon planning position SpatialWorld as a rigorous testbed for future spatial agents.</p> + https://arxiv.org/abs/2606.09669 + Mon, 08 Jun 2026 00:00:00 +0000 + + + FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention + https://arxiv.org/abs/2606.09079 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09079.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yan Wang, Qifan Zhang, Jiachen Yu, Tian Liang, Dongyang Ma, Xiang Hu, Zibo Lin, Chunyang Li, Zhichao Wang, Jia Li, Yujiu Yang, Haitao Mi, Dong Yu</p><p><b>Upvotes:</b> 42</p><p><b>Summary:</b> Conventional LLMs keep the full KV cache loaded during decoding, causing a severe GPU memory bottleneck for ultra-long context serving. In this report, we propose Lookahead Sparse Attention (LSA), a novel inference paradigm powered by a Neural Memory Indexer built upon the DeepSeek-V4 architecture. Rather than passively attending to all historical tokens, LSA proactively predicts future context demands and preserves only the query-critical KV chunks in the GPU memory. Crucially, we instantiate this architecture via a backbone-free decoupled training strategy. By formulating the indexer as a standard dual-encoder architecture, we train it independently using standard retrieval training frameworks without ever loading the massive backbone model into GPU memory. We demonstrate that this "less is more" paradigm significantly maximizes serving efficiency while acting as an effective attention denoiser in tasks that rely on long-term global memory. Across primary long-context evaluation suites (e.g., LongBench-v2, LongMemEval, and RULER), FM-DS-V4 compresses the average physical KV cache footprint down to merely 13.5% of the full-context baseline, while consistently preserving or slightly elevating downstream accuracy (+0.6% absolute margin on average). Crucially, at extreme 500K scales, FlashMemory suppresses the physical KV cache overhead by over 90% without destabilizing the backbone's core reasoning capacities.</p> + https://arxiv.org/abs/2606.09079 + Mon, 08 Jun 2026 00:00:00 +0000 + + + Latent Spatial Memory for Video World Models + https://arxiv.org/abs/2606.09828 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09828.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Weijie Wang, Haoyu Zhao, Yifan Yang, Feng Chen, Zeyu Zhang, Yefei He, Zicheng Duan, Donny Y. Chen, Yuqing Yang, Bohan Zhuang</p><p><b>Upvotes:</b> 47</p><p><b>Summary:</b> Video world models that maintain 3D spatial consistency across generated frames typically rely on explicit point cloud memory constructed in RGB space. This design is both computationally expensive, requiring repeated rendering and VAE encoding, and inherently lossy, as the round trip through pixel space discards rich features of the learned latent representation. In this paper, we introduce latent spatial memory for video world models, a persistent 3D cache that stores scene information directly in the diffusion latent space, avoiding pixel-space reconstruction. Building on this, we propose Mirage, a latent-space spatial memory framework that constructs the memory by lifting latent tokens into 3D via depth-guided back-projection and queries it by synthesizing novel views through direct latent-space warping. This unified formulation eliminates both the information loss of pixel-space reconstruction and the computational burden of repeated encoding and rendering. Experiments show that latent spatial memory achieves up to 10.57times faster end-to-end video generation and 55times reduction in memory footprint relative to explicit 3D baselines. Leveraging the geometric prior of the diffusion model, Mirage attains state-of-the-art performance on WorldScore and strong reconstruction quality on RealEstate10K.</p> + https://arxiv.org/abs/2606.09828 + Mon, 08 Jun 2026 00:00:00 +0000 + + + Hardening Agent Benchmarks with Adversarial Hacker-Fixer Loops + https://arxiv.org/abs/2606.08960 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.08960.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Ziqian Zhong, Ivgeni Segal, Ivan Bercovich, Shashwat Saxena, Kexun Zhang, Aditi Raghunathan</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Agent benchmarks score submissions with outcome verifiers that are typically hand-written and brittle, leaving them open to reward hacking. We audit 1,968 tasks across five terminal-agent benchmarks and find 323 (16%) hackable by frontier models given only the task description. This corrupts both leaderboard rankings and RL training signal, yet the standard response is manual and reactive. We introduce the hacker-fixer loop, a method for building exploit-resistant verifiers without per-task manual patching. The loop alternates three LLM agents: a hacker tries to pass the verifier without solving the task, a fixer patches the verifier to reject each discovered exploit, and a solver confirms the patched verifier still admits legitimate solutions. The loop iterates: each patch reshapes what the verifier rewards, surfacing the next exploit. We further add verifier access, and let patches transfer across tasks, to broaden the exploits the loop discovers. On KernelBench, the loop drives the attack success rate from 62% to 0% on a held-out corpus of publicly reported exploits. We also find that weaker agents in the loop can defend against much stronger hackers: Gemini 3 Flash's loop drives the stronger Gemini 3.1 Pro and Claude Opus 4.7's attack success rate from 76% and 61% to 0% on KernelBench, and Gemini 3.1 Pro's from 39% to 17% on Terminal Bench across 77 tasks. We release Terminal Wrench (323 hackable environments, 3,632 hack trajectories) as a snapshot of the current attack surface, our patched verifiers, the exploits the loop discovered, and our implementation as a basis for future work.</p> + https://arxiv.org/abs/2606.08960 + Mon, 08 Jun 2026 03:00:56 +0000 + + + Experience Makes Skillful: Enabling Generalizable Medical Agent Reasoning via Self-Evolving Skill Memory + https://arxiv.org/abs/2606.09365 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09365.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Haoran Sun, Wenjie Li, Yujie Zhang, Zekai Lin, Fanrui Zhang, Kaitao Chen, Xingqi He, Yichen Li, Mianxin Liu, Lei Liu, Yankai Jiang</p><p><b>Upvotes:</b> 1</p><p><b>Summary:</b> Medical agent systems are increasingly expected to support interactive clinical decision making rather than only static question answering. In such settings, effective agents must reuse prior experience across evolving cases, yet existing memory mechanisms often retain raw historical traces that are redundant, noisy, and difficult to govern. More importantly, they rarely distinguish which memories are truly useful for future reasoning. This limits their ability to accumulate compact and reliable experience for long-horizon clinical reasoning. To close this gap, we propose SkeMex, a post-deployment self-evolution framework that improves medical agents through a skill-based memory without updating model weights. SkeMex distills informative interaction trajectories into structured skills that encode reusable procedural knowledge, and organizes them into a multi-branch repository spanning general, task-specific, and action-level experience. To determine which memories should be reused and retained, SkeMex estimates context-dependent utility from environment feedback and uses it to guide value-aware retrieval and repository governance. A closed-loop ``Read--Write--Assess--Govern" lifecycle further supports continual evolution by writing new skills, updating utilities, promoting useful memories, and removing harmful entries. Experiments across diverse clinical tasks show that SkeMex consistently outperforms representative memory-based agents in both offline and online settings. It also generalizes across model backbones and supports transferable skill memory. All data and code will be released publicly.</p> + https://arxiv.org/abs/2606.09365 + Mon, 08 Jun 2026 11:37:01 +0000 + + + Reasoning Arena: Trace Tournaments When Verifiable Rewards Fall Short + https://arxiv.org/abs/2606.09380 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09380.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Han Zhou, Adam X. Yang, Laurence Aitchison, Anna Korhonen, Albert Q. Jiang</p><p><b>Upvotes:</b> 7</p><p><b>Summary:</b> Reinforcement learning with verifiable rewards (RLVR) has become a leading paradigm for improving the reasoning ability of large language models through outcome-based supervision. However, verifiable rewards frequently become uninformative at the group level: when all sampled traces of a given prompt receive identical rewards, group-relative advantage estimation provides no gradient signal, even though the traces may differ substantially in reasoning quality. We propose Reasoning Arena, an adaptive training framework that routes such non-diverse reward groups to a judge system instead of discarding them. Beyond examining the final answer, Reasoning Arena constructs trace tournaments, where reasoning traces are compared head-to-head to expose finer-grained preferences within the group, converting reasoning quality into rich relative reward signals. To make reward estimation efficient, rather than exhaustively comparing every pair, each new trace is evaluated against a small, dynamically updated pool of previously generated traces as anchors to efficiently establish a relative ranking. We then fit a Bradley-Terry model on the incomplete comparison graph, enabling scalable RL integration without quadratic pairwise comparisons. Empirical results demonstrate that Reasoning Arena consistently outperforms the RLVR baseline by 7.6% on average in competition mathematics and coding benchmarks. By converting otherwise wasted zero-advantage samples into useful gradient updates, our method accelerates training by 27% to 41%, saving nearly 50% of generation compute, and substantially improves overall reasoning performance.</p> + https://arxiv.org/abs/2606.09380 + Mon, 08 Jun 2026 11:57:17 +0000 + + + Optical Reasoning: Rethinking Images as an Expressive Reasoning Medium Beyond Text + https://arxiv.org/abs/2606.09585 + <p><img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2606.09585.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /></p><p><b>Authors:</b> Yutong Bian, Dongjie Cheng, Heming Xia, Yongqi Li, Wenjie Li</p><p><b>Upvotes:</b> 2</p><p><b>Summary:</b> Chain-of-Thought (CoT) improves the performance of Large Language Models (LLMs) and has been extended to Multimodal Large Language Models (MLLMs). More recent work further moves from text-based multimodal reasoning toward interleaved-modal reasoning, where intermediate steps can incorporate both textual rationales and visual evidence. In this work, we propose a bolder and more ambitious idea: could images alone serve as the reasoning medium for both language and multimodal tasks? To explore this, we propose optical reasoning, which treats images as a standalone reasoning medium. We instantiate this concept with two variants: typographic-based optical reasoning, which optimizes visual layouts for compact rationale rendering, and graphical-based optical reasoning, which composes text and graphical elements into structured visual rationales. Across mathematical, scientific, and interleaved-modal reasoning benchmarks, optical reasoning can match or even exceed traditional text reasoning while reducing reasoning tokens by an average of 28.57% on language tasks and 16% on multimodal tasks, achieving 1.96 times the token efficiency of text reasoning. These results show that images can effectively and efficiently encode rationales while providing a unified visual canvas for reasoning.</p> + https://arxiv.org/abs/2606.09585 + Mon, 08 Jun 2026 14:58:59 +0000