ずんだもんのHugging Faceニュース

【本日の論文】1. VideoChat3: Fully Open Video MLLM for Efficient and Generalist Video Understanding https://huggingface.co/papers/2607.149352. SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning https://huggingface.co/papers/2607.147773. SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration https://huggingface.co/papers/2607.152574. LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget https://huggingface.co/papers/2607.149525. BadWAM: When World-Action Models Dream Right but Act Wrong https://huggingface.co/papers/2607.15207

【本日の論文】1. Harness Handbook: Making Evolving Agent Harnesses Readable,Navigable, and Editable https://huggingface.co/papers/2607.132852. Boogu-Image-0.1: Boosting Open-Source Unified Multimodal Understanding and Generation https://huggingface.co/papers/2607.131253. Ring-Zero: Scaling Zero RL to a Trillion Parameters for Emergent Reasoning https://huggingface.co/papers/2607.123954. KnowAct-GUIClaw: Know Deeply, Act Perfectly, Personal GUI Assistant with Self-Evolving Memory and Skill https://huggingface.co/papers/2607.126255. OvisOCR2 Technical Report https://huggingface.co/papers/2607.13639

【本日の論文】1. SynthDocBench: Controlled Benchmark for Long-Context Visual Document Understanding https://huggingface.co/papers/2607.104002. Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image Generation https://huggingface.co/papers/2607.118863. Search Beyond What Can Be Taught: Evolving the Knowledge Boundary in Agentic Visual Generation https://huggingface.co/papers/2607.053824. Blind-Spots-Bench: Evaluating Blind Spots in Multimodal Models https://huggingface.co/papers/2607.083175. Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models https://huggingface.co/papers/2607.12463

【本日の論文】1. Weak-to-Strong Generalization via Direct On-Policy Distillation https://huggingface.co/papers/2607.053942. ABot-N1: Toward a General Visual Language Navigation Foundation Model https://huggingface.co/papers/2607.103833. ABot-AgentOS: A General Robotic Agent OS with Lifelong Multi-modal Memory https://huggingface.co/papers/2607.103504. 4D Human-Scene Reconstruction from Low-Overlap Captures https://huggingface.co/papers/2607.091255. LightMem-Ego: Your AI Memory for Everyday Life https://huggingface.co/papers/2607.11487

【本日の論文】1. Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks with Dense Reward-Based Grading https://huggingface.co/papers/2607.089642. Scalable Visual Pretraining for Language Intelligence https://huggingface.co/papers/2607.096573. Video Generation Models are General-Purpose Vision Learners https://huggingface.co/papers/2607.090244. Trust Region Policy Distillation https://huggingface.co/papers/2607.047515. KronQ: LLM Quantization via Kronecker-Factored Hessian https://huggingface.co/papers/2607.07964

【本日の論文】1. Vidu S1: A Real-Time Interactive Video Generation Model https://huggingface.co/papers/2607.031182. Video-Oasis: Rethinking Evaluation of Video Understanding https://huggingface.co/papers/2603.296163. Why Can't I Open My Drawer? Mitigating Object-Driven Shortcuts in Zero-Shot Compositional Action Recognition https://huggingface.co/papers/2601.162114. Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation https://huggingface.co/papers/2607.087585. LongE2V: Long-Horizon Event-based Video Reconstruction, Prediction, and Frame Interpolation with Video Diffusion Models https://huggingface.co/papers/2607.08770

【本日の論文】1. Vidu S1: A Real-Time Interactive Video Generation Model https://huggingface.co/papers/2607.031182. Video-Oasis: Rethinking Evaluation of Video Understanding https://huggingface.co/papers/2603.296163. Why Can't I Open My Drawer? Mitigating Object-Driven Shortcuts in Zero-Shot Compositional Action Recognition https://huggingface.co/papers/2601.162114. Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation https://huggingface.co/papers/2607.087585. UniClawBench: A Universal Benchmark for Proactive Agents on Real-World Tasks https://huggingface.co/papers/2607.08768

【本日の論文】1. Vidu S1: A Real-Time Interactive Video Generation Model https://huggingface.co/papers/2607.031182. Video-Oasis: Rethinking Evaluation of Video Understanding https://huggingface.co/papers/2603.296163. Why Can't I Open My Drawer? Mitigating Object-Driven Shortcuts in Zero-Shot Compositional Action Recognition https://huggingface.co/papers/2601.162114. UniClawBench: A Universal Benchmark for Proactive Agents on Real-World Tasks https://huggingface.co/papers/2607.087685. Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation https://huggingface.co/papers/2607.08758

【本日の論文】1. Accurate, Interdisciplinary and Transparent Structure-property Understanding with Deep Native Structural Reasoning https://huggingface.co/papers/2607.077082. Dual Latent Memory in Vision-Language-Action Models for Robotic Manipulation https://huggingface.co/papers/2607.076083. Scaling Mixture-of-Experts Video Pretraining for Embodied Intelligence https://huggingface.co/papers/2607.076754. Infinite Worlds with Versatile Interactions https://huggingface.co/papers/2607.075345. Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning https://huggingface.co/papers/2607.07508

【本日の論文】1. RynnWorld-4D: 4D Embodied World Models for Robotic Manipulation https://huggingface.co/papers/2607.065592. AlayaWorld: Long-Horizon and Playable Video World Generation https://huggingface.co/papers/2607.062913. RynnWorld-Teleop: An Action-Conditioned World Model for Digital Teleoperation https://huggingface.co/papers/2607.065584. Hierarchical Sparse Attention Done Right: Toward Infinite Context Modeling https://huggingface.co/papers/2607.029805. Vision as Unified Multimodal Generation https://huggingface.co/papers/2607.06560

【本日の論文】1. OmniOpt: Taxonomy, Geometry, and Benchmarking of Modern Optimizers https://huggingface.co/papers/2607.040332. UI-MOPD: Multi-Platform On-Policy Distillation for Continual GUI Agent Learning https://huggingface.co/papers/2607.044253. ResearchStudio-Reel: Automate the Last Mile of Research from Paper to Poster, Video, and Blog https://huggingface.co/papers/2607.044384. PixWorld: Unifying 3D Scene Generation and Reconstruction in Pixel Space https://huggingface.co/papers/2607.053735. ResearchStudio-Idea: An Evidence-Grounded Research-Ideation Skill Suite from ML Conference Outcomes https://huggingface.co/papers/2607.04439

【本日の論文】1. The Mirage of Optimizing Training Policies: Monotonic Inference Policies as the Real Objective for LLM Reinforcement Learning https://huggingface.co/papers/2606.295262. Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots https://huggingface.co/papers/2607.025013. OrbitQuant: Data-Agnostic Quantization for Image and Video Diffusion Transformers https://huggingface.co/papers/2607.024614. VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon https://huggingface.co/papers/2607.018045. DataComp-VLM: Improved Open Datasets for Vision-Language Models https://huggingface.co/papers/2606.28551

【本日の論文】1. Program-as-Weights: A Programming Paradigm for Fuzzy Functions https://huggingface.co/papers/2607.025122. AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents https://huggingface.co/papers/2607.022553. EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments https://huggingface.co/papers/2607.024404. Morphing into Hybrid Attention Models https://huggingface.co/papers/2606.305625. Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling https://huggingface.co/papers/2607.01642

【本日の論文】1. Program-as-Weights: A Programming Paradigm for Fuzzy Functions https://huggingface.co/papers/2607.025122. AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents https://huggingface.co/papers/2607.022553. EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments https://huggingface.co/papers/2607.024404. Morphing into Hybrid Attention Models https://huggingface.co/papers/2606.305625. Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling https://huggingface.co/papers/2607.01642

【本日の論文】1. Program-as-Weights: A Programming Paradigm for Fuzzy Functions https://huggingface.co/papers/2607.025122. EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments https://huggingface.co/papers/2607.024403. AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents https://huggingface.co/papers/2607.022554. Morphing into Hybrid Attention Models https://huggingface.co/papers/2606.305625. Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling https://huggingface.co/papers/2607.01642

【本日の論文】1. PerceptionRubrics: Calibrating Multimodal Evaluation to Human Perception https://huggingface.co/papers/2606.283222. ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving https://huggingface.co/papers/2607.004663. TurboServe: Serving Streaming Video Generation Efficiently and Economically https://huggingface.co/papers/2606.192714. MemSyco-Bench: Benchmarking Sycophancy in Agent Memory https://huggingface.co/papers/2607.010715. Multimodal Continuous Reasoning via Asymmetric Mutual Variational Learning https://huggingface.co/papers/2607.00461

【本日の論文】1. Orca: The World is in Your Mind https://huggingface.co/papers/2606.305342. Dockerless: Environment-Free Program Verifier for Coding Agents https://huggingface.co/papers/2606.284363. DOPD: Dual On-policy Distillation https://huggingface.co/papers/2606.306264. BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding https://huggingface.co/papers/2606.313155. Does VLA Even Know the Basics? Measuring Commonsense and World Knowledge Retention in Vision-Language-Action Models https://huggingface.co/papers/2606.19297

【本日の論文】1. LoopCoder-v2: Only Loop Once for Efficient Test-Time Computation Scaling https://huggingface.co/papers/2606.180232. Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradients https://huggingface.co/papers/2606.182163. ACE-Ego-0: Unifying Egocentric Human and Robotic Data for VLA Pretraining https://huggingface.co/papers/2606.172004. GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine? https://huggingface.co/papers/2606.178615. LectūraAgents: A Multi-Agent Framework for Adaptive Personalized AI-Assisted Learning and Embodied Teaching https://huggingface.co/papers/2606.16428

【本日の論文】1. JoyAI-VL-Interaction: Real-Time Vision-Language Interaction Intelligence https://huggingface.co/papers/2606.147772. Data Journalist Agent: Transforming Data into Verifiable Multimodal Stories https://huggingface.co/papers/2606.111763. Geometric Action Model for Robot Policy Learning https://huggingface.co/papers/2606.170464. DreamX-World 1.0: A General-Purpose Interactive World Model https://huggingface.co/papers/2606.169935. FastContext: Training Efficient Repository Explorer for Coding Agents https://huggingface.co/papers/2606.14066

【本日の論文】1. OmniDirector: General Multi-Shot Camera Cloning without Cross-Paired Data https://huggingface.co/papers/2606.134322. APPO: Agentic Procedural Policy Optimization https://huggingface.co/papers/2606.123843. Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents https://huggingface.co/papers/2606.060364. From Chatbot to Digital Colleague: The Paradigm Shift Toward Persistent Autonomous AI https://huggingface.co/papers/2606.145025. Orchestra-o1: Omnimodal Agent Orchestration https://huggingface.co/papers/2606.13707

【本日の論文】1. EvoArena: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments https://huggingface.co/papers/2606.136812. MiniMax Sparse Attention https://huggingface.co/papers/2606.133923. WeaveBench: A Long-Horizon, Real-World Benchmark for Computer-Use Agents with Hybrid Interfaces https://huggingface.co/papers/2606.094264. SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning https://huggingface.co/papers/2606.136735. InterleaveThinker: Reinforcing Agentic Interleaved Generation https://huggingface.co/papers/2606.13679

【本日の論文】1. EvoArena: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments https://huggingface.co/papers/2606.136812. MiniMax Sparse Attention https://huggingface.co/papers/2606.133923. WeaveBench: A Long-Horizon, Real-World Benchmark for Computer-Use Agents with Hybrid Interfaces https://huggingface.co/papers/2606.094264. SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning https://huggingface.co/papers/2606.136735. InterleaveThinker: Reinforcing Agentic Interleaved Generation https://huggingface.co/papers/2606.13679

【本日の論文】1. EvoArena: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments https://huggingface.co/papers/2606.136812. SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning https://huggingface.co/papers/2606.136733. MiniMax Sparse Attention https://huggingface.co/papers/2606.133924. InterleaveThinker: Reinforcing Agentic Interleaved Generation https://huggingface.co/papers/2606.136795. Robust-U1: Can MLLMs Self-Recover Corrupted Visual Content for Robust Understanding? https://huggingface.co/papers/2606.08063

【本日の論文】1. Redesign Mixture-of-Experts Routers with Manifold Power Iteration https://huggingface.co/papers/2606.123972. Toward Generalist Autonomous Research via Hypothesis-Tree Refinement https://huggingface.co/papers/2606.119263. Agentic Environment Engineering for Large Language Models: A Survey of Environment Modeling, Synthesis, Evaluation, and Application https://huggingface.co/papers/2606.121914. Claw-SWE-Bench: A Benchmark for Evaluating OpenClaw-style Agent Harnesses on Coding Tasks https://huggingface.co/papers/2606.123445. Beyond Scalar Rewards by Internalizing Reasoning into Score Distributions https://huggingface.co/papers/2606.09076

【本日の論文】1. ABot-Earth 0.5: Generative 3D Earth Model https://huggingface.co/papers/2606.099672. Kwai Keye-VL-2.0 Technical Report https://huggingface.co/papers/2606.106513. Role-Agent: Bootstrapping LLM Agents via Dual-Role Evolution https://huggingface.co/papers/2606.109174. Retrospective Harness Optimization: Improving LLM Agents via Self-Preference over Trajectory Rollouts https://huggingface.co/papers/2606.059225. SearchSwarm: Towards Delegation Intelligence in Agentic LLMs for Long-Horizon Deep Research https://huggingface.co/papers/2606.09730

【本日の論文】1. SWE-Explore: Benchmarking How Coding Agents Explore Repositories https://huggingface.co/papers/2606.072972. On the Geometry of On-Policy Distillation https://huggingface.co/papers/2606.070823. Agents' Last Exam https://huggingface.co/papers/2606.054054. LatentSkill: From In-Context Textual Skills to In-Weight Latent Skills for LLM Agents https://huggingface.co/papers/2606.060875. Latent Spatial Memory for Video World Models https://huggingface.co/papers/2606.09828

【本日の論文】1. Your UnEmbedding Matrix is Secretly a Feature Lens for Text Embeddings https://huggingface.co/papers/2606.075022. SoCRATES: Towards Reliable Automated Evaluation of Proactive LLM Mediation across Domains and Socio-cognitive Variations https://huggingface.co/papers/2606.055633. GENEB: Why Genomic Models Are Hard to Compare https://huggingface.co/papers/2606.045254. MMAE: A Massive Multitask Audio Editing Benchmark https://huggingface.co/papers/2606.072295. AnchorWorld: Embodied Egocentric World Simulation with View-based Evolution Customization https://huggingface.co/papers/2606.07326

【本日の論文】1. Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution https://huggingface.co/papers/2606.064922. ArcANE: Do Role-Playing Language Agents Stay in Character at the Right Time? https://huggingface.co/papers/2606.055533. TIDE: Proactive Multi-Problem Discovery via Template-Guided Iteration https://huggingface.co/papers/2606.047434. AdaPlanBench: Evaluating Adaptive Planning in Large Language Model Agents under World and User Constraints https://huggingface.co/papers/2606.056225. VideoKR: Towards Knowledge- and Reasoning-Intensive Video Understanding https://huggingface.co/papers/2606.05259

【本日の論文】1. Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution https://huggingface.co/papers/2606.064922. ArcANE: Do Role-Playing Language Agents Stay in Character at the Right Time? https://huggingface.co/papers/2606.055533. TIDE: Proactive Multi-Problem Discovery via Template-Guided Iteration https://huggingface.co/papers/2606.047434. AdaPlanBench: Evaluating Adaptive Planning in Large Language Model Agents under World and User Constraints https://huggingface.co/papers/2606.056225. VideoKR: Towards Knowledge- and Reasoning-Intensive Video Understanding https://huggingface.co/papers/2606.05259

【本日の論文】1. Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution https://huggingface.co/papers/2606.064922. ArcANE: Do Role-Playing Language Agents Stay in Character at the Right Time? https://huggingface.co/papers/2606.055533. TIDE: Proactive Multi-Problem Discovery via Template-Guided Iteration https://huggingface.co/papers/2606.047434. AdaPlanBench: Evaluating Adaptive Planning in Large Language Model Agents under World and User Constraints https://huggingface.co/papers/2606.056225. VideoKR: Towards Knowledge- and Reasoning-Intensive Video Understanding https://huggingface.co/papers/2606.05259

【本日の論文】1. Audio Interaction Model https://huggingface.co/papers/2606.051212. Cosmos 3: Omnimodal World Models for Physical AI https://huggingface.co/papers/2606.028003. Where Do Deep-Research Agents Go Wrong? Span-Level Error Localization in Agent Trajectories https://huggingface.co/papers/2606.020604. Reproducing, Analyzing, and Detecting Reward Hacking in Rubric-Based Reinforcement Learning https://huggingface.co/papers/2606.049235. OVO-S-Bench: A Hierarchical Benchmark for Streaming Spatial Intelligence in Multimodal LLMs https://huggingface.co/papers/2606.03890

【本日の論文】1. OCC-RAG: Optimal Cognitive Core for Faithful Question Answering https://huggingface.co/papers/2606.006832. From Activation to Causality: Discovery of Causal Visual Representations in the Human Brain https://huggingface.co/papers/2605.238953. Trust Region On-Policy Distillation https://huggingface.co/papers/2606.012494. Humanoid-GPT: Scaling Data and Structure for Zero-Shot Motion Tracking https://huggingface.co/papers/2606.039855. KVarN: Variance-Normalized KV-Cache Quantization Mitigates Error Accumulation in Reasoning Tasks https://huggingface.co/papers/2606.03458

【本日の論文】1. Crafter: A Multi-Agent Harness for Editable Scientific Figure Generation from Diverse Inputs https://huggingface.co/papers/2605.306112. On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters https://huggingface.co/papers/2606.024373. A Matter of TASTE: Improving Coverage and Difficulty of Agent Benchmarks https://huggingface.co/papers/2605.285564. K-BrowseComp: A Web Browsing Agent Benchmark Grounded in Korean Contexts https://huggingface.co/papers/2606.024045. Harness-1: Reinforcement Learning for Search Agents with State-Externalizing Harnesses https://huggingface.co/papers/2606.02373

【本日の論文】1. GrepSeek: Training Search Agents for Direct Corpus Interaction https://huggingface.co/papers/2605.293072. COLLEAGUE.SKILL: Automated AI Skill Generation via Expert Knowledge Distillation https://huggingface.co/papers/2605.312643. Trust-Region Behavior Blending for On-Policy Distillation https://huggingface.co/papers/2605.311594. Representation Forcing for Bottleneck-Free Unified Multimodal Models https://huggingface.co/papers/2605.316045. SwanVoice: Expressive Long-Form Zero-Shot Speech Synthesis for Both Monologue and Dialogue https://huggingface.co/papers/2605.30993

【本日の論文】1. AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security https://huggingface.co/papers/2605.298012. Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments https://huggingface.co/papers/2605.302803. OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge Sources https://huggingface.co/papers/2605.292504. CollectionLoRA: Collecting 50 Effects in 1 LoRA via Multi-Teacher On-Policy Distillation https://huggingface.co/papers/2605.253785. minWM: A Full-Stack Open-Source Framework for Real-Time Interactive Video World Models https://huggingface.co/papers/2605.30263

【本日の論文】1. AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security https://huggingface.co/papers/2605.298012. Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments https://huggingface.co/papers/2605.302803. OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge Sources https://huggingface.co/papers/2605.292504. CollectionLoRA: Collecting 50 Effects in 1 LoRA via Multi-Teacher On-Policy Distillation https://huggingface.co/papers/2605.253785. minWM: A Full-Stack Open-Source Framework for Real-Time Interactive Video World Models https://huggingface.co/papers/2605.30263

【本日の論文】1. AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security https://huggingface.co/papers/2605.298012. Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments https://huggingface.co/papers/2605.302803. OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge Sources https://huggingface.co/papers/2605.292504. CollectionLoRA: Collecting 50 Effects in 1 LoRA via Multi-Teacher On-Policy Distillation https://huggingface.co/papers/2605.253785. minWM: A Full-Stack Open-Source Framework for Real-Time Interactive Video World Models https://huggingface.co/papers/2605.30263

【本日の論文】1. Gamma-World: Generative Multi-Agent World Modeling Beyond Two Players https://huggingface.co/papers/2605.288162. ProRL: Effective Reinforcement Learning for Proactive Recommendation via Rectified Policy Gradient Estimation https://huggingface.co/papers/2605.282933. Agent Explorative Policy Optimization for Multimodal Agentic Reasoning https://huggingface.co/papers/2605.287744. From Pixels to Words -- Towards Native One-Vision Models at Scale https://huggingface.co/papers/2605.288205. Self-Improving Language Models with Bidirectional Evolutionary Search https://huggingface.co/papers/2605.28814

【本日の論文】1. LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decoding https://huggingface.co/papers/2605.273652. EvalVerse: Pipeline-Aware and Expert-Calibrated Benchmarking for Professional Cinematic Video Generation https://huggingface.co/papers/2605.232713. SpatialBench: Is Your Spatial Foundation Model an All-Round Player? https://huggingface.co/papers/2605.273674. MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research https://huggingface.co/papers/2605.261145. Geometry-Aware Representation Denoising for Robust Multi-view 3D Reconstruction https://huggingface.co/papers/2605.26230

【本日の論文】1. DVAO: Dynamic Variance-adaptive Advantage Optimization for Multi-reward Reinforcement Learning https://huggingface.co/papers/2605.256042. WBench: A Comprehensive Multi-turn Benchmark for Interactive Video World Model Evaluation https://huggingface.co/papers/2605.258743. Macaron-A2UI: A Model for Generative UI in Personal Agents https://huggingface.co/papers/2605.248304. Foundation Protocol: A Coordination Layer for Agentic Society https://huggingface.co/papers/2605.232185. TriSplat: Simulation-Ready Feed-Forward 3D Scene Reconstruction https://huggingface.co/papers/2605.26115

【本日の論文】1. SkillOpt: Executive Strategy for Self-Evolving Agent Skills https://huggingface.co/papers/2605.239042. Rethinking Cross-Layer Information Routing in Diffusion Transformers https://huggingface.co/papers/2605.207083. Lens: Rethinking Training Efficiency for Foundational Text-to-Image Models https://huggingface.co/papers/2605.215734. SciAtlas: A Large-Scale Knowledge Graph for Automated Scientific Research https://huggingface.co/papers/2605.228785. StepAudio 2.5 Technical Report https://huggingface.co/papers/2605.23463

【本日の論文】1. DelTA: Discriminative Token Credit Assignment for Reinforcement Learning from Verifiable Rewards https://huggingface.co/papers/2605.214672. TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation https://huggingface.co/papers/2605.223553. Perception or Prejudice: Can MLLMs Go Beyond First Impressions of Personality? https://huggingface.co/papers/2605.221094. π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows https://huggingface.co/papers/2605.146785. Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps https://huggingface.co/papers/2605.16928

【本日の論文】1. DelTA: Discriminative Token Credit Assignment for Reinforcement Learning from Verifiable Rewards https://huggingface.co/papers/2605.214672. TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation https://huggingface.co/papers/2605.223553. Perception or Prejudice: Can MLLMs Go Beyond First Impressions of Personality? https://huggingface.co/papers/2605.221094. π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows https://huggingface.co/papers/2605.146785. Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps https://huggingface.co/papers/2605.16928

【本日の論文】1. TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation https://huggingface.co/papers/2605.223552. Perception or Prejudice: Can MLLMs Go Beyond First Impressions of Personality? https://huggingface.co/papers/2605.221093. DelTA: Discriminative Token Credit Assignment for Reinforcement Learning from Verifiable Rewards https://huggingface.co/papers/2605.214674. π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows https://huggingface.co/papers/2605.146785. Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps https://huggingface.co/papers/2605.16928

【本日の論文】1. Mega-ASR: Towards In-the-wild^2 Speech Recognition via Scaling up Real-world Acoustic Simulation https://huggingface.co/papers/2605.198332. Video2GUI: Synthesizing Large-Scale Interaction Trajectories for Generalized GUI Agent Pretraining https://huggingface.co/papers/2605.147473. Enhancing Train-Free Infinite-Frame Generation for Consistent Long Videos https://huggingface.co/papers/2605.182334. IndusAgent: Reinforcing Open-Vocabulary Industrial Anomaly Detection with Agentic Tools https://huggingface.co/papers/2605.206825. You Only Need Minimal RLVR Training: Extrapolating LLMs via Rank-1 Trajectories https://huggingface.co/papers/2605.21468

【本日の論文】1. When Vision Speaks for Sound https://huggingface.co/papers/2605.164032. Active Learners as Efficient PRP Rerankers https://huggingface.co/papers/2605.142363. Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information https://huggingface.co/papers/2605.116094. GoLongRL: Capability-Oriented Long Context Reinforcement Learning with Multitask Alignment https://huggingface.co/papers/2605.195775. OpenComputer: Verifiable Software Worlds for Computer-Use Agents https://huggingface.co/papers/2605.19769

【本日の論文】1. Code as Agent Harness https://huggingface.co/papers/2605.187472. SkillsVote: Lifecycle Governance of Agent Skills from Collection, Recommendation to Evolution https://huggingface.co/papers/2605.184013. LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generation https://huggingface.co/papers/2605.187394. Lance: Unified Multimodal Modeling by Multi-Task Synergy https://huggingface.co/papers/2605.186785. AI for Auto-Research: Roadmap & User Guide https://huggingface.co/papers/2605.18661

【本日の論文】1. CiteVQA: Benchmarking Evidence Attribution for Trustworthy Document Intelligence https://huggingface.co/papers/2605.128822. PhysBrain 1.0 Technical Report https://huggingface.co/papers/2605.152983. MMSkills: Towards Multimodal Skills for General Visual Agents https://huggingface.co/papers/2605.135274. FashionChameleon: Towards Real-Time and Interactive Human-Garment Video Customization https://huggingface.co/papers/2605.158245. Learning to Foresee: Unveiling the Unlocking Efficiency of On-Policy Distillation https://huggingface.co/papers/2605.11739

【本日の論文】1. Achieving Gold-Medal-Level Olympiad Reasoning via Simple and Unified Scaling https://huggingface.co/papers/2605.133012. Causal Forcing++: Scalable Few-Step Autoregressive Diffusion Distillation for Real-Time Interactive Video Generation https://huggingface.co/papers/2605.151413. Self-Distilled Agentic Reinforcement Learning https://huggingface.co/papers/2605.151554. MemLens: Benchmarking Multimodal Long-Term Memory in Large Vision-Language Models https://huggingface.co/papers/2605.149065. SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion Transformer https://huggingface.co/papers/2605.15178

【本日の論文】1. Achieving Gold-Medal-Level Olympiad Reasoning via Simple and Unified Scaling https://huggingface.co/papers/2605.133012. Causal Forcing++: Scalable Few-Step Autoregressive Diffusion Distillation for Real-Time Interactive Video Generation https://huggingface.co/papers/2605.151413. Self-Distilled Agentic Reinforcement Learning https://huggingface.co/papers/2605.151554. MemLens: Benchmarking Multimodal Long-Term Memory in Large Vision-Language Models https://huggingface.co/papers/2605.149065. SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion Transformer https://huggingface.co/papers/2605.15178