ずんだもんのHugging Faceニュース

【本日の論文】 1. MetaClaw: Just Talk -- An Agent That Meta-Learns and Evolves in the Wild https://huggingface.co/papers/2603.17187 2. Video-CoE: Reinforcing Video Event Prediction via Chain of Events https://huggingface.co/papers/2603.14935 3. MosaicMem: Hybrid Spatial Memory for Controllable Video World Models https://huggingface.co/papers/2603.17117 4. Alignment Makes Language Models Normative,...

【本日の論文】 1. InCoder-32B: Code Foundation Model for Industrial Scenarios https://huggingface.co/papers/2603.16790 2. MiroThinker-1.7 & H1: Towards Heavy-Duty Research Agents via Verification https://huggingface.co/papers/2603.15726 3. Qianfan-OCR: A Unified End-to-End Model for Document Intelligence https://huggingface.co/papers/2603.13398 4. Kinema4D: Kinematic 4D World Modeling for Spatiote...

【本日の論文】 1. AI Can Learn Scientific Taste https://huggingface.co/papers/2603.14473 2. OpenSeeker: Democratizing Frontier Search Agents by Fully Open-Sourcing Training Data https://huggingface.co/papers/2603.15594 3. EnterpriseOps-Gym: Environments and Evaluations for Stateful Agentic Planning and Tool Use in Enterprise Settings https://huggingface.co/papers/2603.13594 4. HSImul3R: Physics-in...

【本日の論文】 1. LMEB: Long-horizon Memory Embedding Benchmark https://huggingface.co/papers/2603.12572 2. Cheers: Decoupling Patch Details from Semantic Representations Enables Unified Multimodal Comprehension and Generation https://huggingface.co/papers/2603.12793 3. Can Vision-Language Models Solve the Shell Game? https://huggingface.co/papers/2603.08436 4. daVinci-Env: Open SWE Environment Sy...

【本日の論文】 1. Spatial-TTT: Streaming Visual-based Spatial Intelligence with Test-Time Training https://huggingface.co/papers/2603.12255 2. Strategic Navigation or Stochastic Search? How Agents and Humans Reason Over Document Collections https://huggingface.co/papers/2603.12180 3. IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse https://huggingface.co/papers/2603.12201 4. V...

【本日の論文】 1. Spatial-TTT: Streaming Visual-based Spatial Intelligence with Test-Time Training https://huggingface.co/papers/2603.12255 2. Strategic Navigation or Stochastic Search? How Agents and Humans Reason Over Document Collections https://huggingface.co/papers/2603.12180 3. IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse https://huggingface.co/papers/2603.12201 4. V...

【本日の論文】 1. Spatial-TTT: Streaming Visual-based Spatial Intelligence with Test-Time Training https://huggingface.co/papers/2603.12255 2. Strategic Navigation or Stochastic Search? How Agents and Humans Reason Over Document Collections https://huggingface.co/papers/2603.12180 3. IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse https://huggingface.co/papers/2603.12201 4. V...

【本日の論文】 1. OpenClaw-RL: Train Any Agent Simply by Talking https://huggingface.co/papers/2603.10165 2. Flash-KMeans: Fast and Memory-Efficient Exact K-Means https://huggingface.co/papers/2603.09229 3. MA-EgoQA: Question Answering over Egocentric Videos from Multiple Embodied Agents https://huggingface.co/papers/2603.09827 4. LLM2Vec-Gen: Generative Embeddings from Large Language Models https...

【本日の論文】 1. Geometry-Guided Reinforcement Learning for Multi-view Consistent 3D Scene Editing https://huggingface.co/papers/2603.03143 2. Thinking to Recall: How Reasoning Unlocks Parametric Knowledge in LLMs https://huggingface.co/papers/2603.09906 3. Omni-Diffusion: Unified Multimodal Understanding and Generation with Masked Discrete Diffusion https://huggingface.co/papers/2603.06577 4. MM...

【本日の論文】 1. Lost in Stories: Consistency Bugs in Long Story Generation by LLMs https://huggingface.co/papers/2603.05890 2. Holi-Spatial: Evolving Video Streams into Holistic 3D Spatial Intelligence https://huggingface.co/papers/2603.07660 3. LoGeR: Long-Context Geometric Reconstruction with Hybrid Memory https://huggingface.co/papers/2603.03269 4. Believe Your Model: Distribution-Guided Conf...

【本日の論文】 1. Penguin-VL: Exploring the Efficiency Limits of VLM with LLM-based Vision Encoders https://huggingface.co/papers/2603.06569 2. BandPO: Bridging Trust Regions and Ratio Clipping via Probability-Aware Bounds for LLM Reinforcement Learning https://huggingface.co/papers/2603.04918 3. Planning in 8 Tokens: A Compact Discrete Tokenizer for Latent World Model https://huggingface.co/paper...

【本日の論文】 1. MOOSE-Star: Unlocking Tractable Training for Scientific Discovery by Breaking the Complexity Barrier https://huggingface.co/papers/2603.03756 2. SkillNet: Create, Evaluate, and Connect AI Skills https://huggingface.co/papers/2603.04448 3. DARE: Aligning LLM Agents with the R Statistical Ecosystem via Distribution-Aware Retrieval https://huggingface.co/papers/2603.04743 4. AgentVi...

【本日の論文】 1. MOOSE-Star: Unlocking Tractable Training for Scientific Discovery by Breaking the Complexity Barrier https://huggingface.co/papers/2603.03756 2. SkillNet: Create, Evaluate, and Connect AI Skills https://huggingface.co/papers/2603.04448 3. DARE: Aligning LLM Agents with the R Statistical Ecosystem via Distribution-Aware Retrieval https://huggingface.co/papers/2603.04743 4. AgentVi...

【本日の論文】 1. MOOSE-Star: Unlocking Tractable Training for Scientific Discovery by Breaking the Complexity Barrier https://huggingface.co/papers/2603.03756 2. SkillNet: Create, Evaluate, and Connect AI Skills https://huggingface.co/papers/2603.04448 3. DARE: Aligning LLM Agents with the R Statistical Ecosystem via Distribution-Aware Retrieval https://huggingface.co/papers/2603.04743 4. RoboPoc...

【本日の論文】 1. Helios: Real Real-Time Long Video Generation Model https://huggingface.co/papers/2603.04379 2. T2S-Bench & Structure-of-Thought: Benchmarking and Prompting Comprehensive Text-to-Structure Reasoning https://huggingface.co/papers/2603.03790 3. Heterogeneous Agent Collaborative Reinforcement Learning https://huggingface.co/papers/2603.02604 4. Proact-VL: A Proactive VideoLLM for Rea...

【本日の論文】 1. Utonia: Toward One Encoder for All Point Clouds https://huggingface.co/papers/2603.03283 2. UniG2U-Bench: Do Unified Models Advance Multimodal Understanding? https://huggingface.co/papers/2603.03241 3. Beyond Language Modeling: An Exploration of Multimodal Pretraining https://huggingface.co/papers/2603.03276 4. BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixi...

【本日の論文】 1. From Scale to Speed: Adaptive Test-Time Scaling for Image Editing https://huggingface.co/papers/2603.00141 2. OmniLottie: Generating Vector Animations via Parameterized Lottie Tokens https://huggingface.co/papers/2603.02138 3. SWE-rebench V2: Language-Agnostic SWE Task Collection at Scale https://huggingface.co/papers/2602.23866 4. RubricBench: Aligning Model-Generated Rubrics wi...

【本日の論文】 1. dLLM: Simple Diffusion Language Modeling https://huggingface.co/papers/2602.22661 2. Enhancing Spatial Understanding in Image Generation via Reward Modeling https://huggingface.co/papers/2602.24233 3. Recovered in Translation: Efficient Pipeline for Automated Translation of Benchmarks and Datasets https://huggingface.co/papers/2602.22207 4. CUDA Agent: Large-Scale Agentic RL for...

【本日の論文】 1. The Trinity of Consistency as a Defining Principle for General World Models https://huggingface.co/papers/2602.23152 2. From Blind Spots to Gains: Diagnostic-Driven Iterative Training for Large Multimodal Models https://huggingface.co/papers/2602.22859 3. MobilityBench: A Benchmark for Evaluating Route-Planning Agents in Real-World Mobility Scenarios https://huggingface.co/papers...

【本日の論文】 1. The Trinity of Consistency as a Defining Principle for General World Models https://huggingface.co/papers/2602.23152 2. From Blind Spots to Gains: Diagnostic-Driven Iterative Training for Large Multimodal Models https://huggingface.co/papers/2602.22859 3. MobilityBench: A Benchmark for Evaluating Route-Planning Agents in Real-World Mobility Scenarios https://huggingface.co/papers...

【本日の論文】 1. The Trinity of Consistency as a Defining Principle for General World Models https://huggingface.co/papers/2602.23152 2. From Blind Spots to Gains: Diagnostic-Driven Iterative Training for Large Multimodal Models https://huggingface.co/papers/2602.22859 3. MobilityBench: A Benchmark for Evaluating Route-Planning Agents in Real-World Mobility Scenarios https://huggingface.co/papers...

【本日の論文】 1. HyTRec: A Hybrid Temporal-Aware Attention Architecture for Long Behavior Sequential Recommendation https://huggingface.co/papers/2602.18283 2. MolHIT: Advancing Molecular-Graph Generation with Hierarchical Discrete Diffusion Models https://huggingface.co/papers/2602.17602 3. DreamID-Omni: Unified Framework for Controllable Human-Centric Audio-Video Generation https://huggingface....

【本日の論文】 1. On Data Engineering for Scaling LLM Terminal Capabilities https://huggingface.co/papers/2602.21193 2. Query-focused and Memory-aware Reranker for Long Context Processing https://huggingface.co/papers/2602.12192 3. PyVision-RL: Forging Open Agentic Vision Models via RL https://huggingface.co/papers/2602.20739 4. From Perception to Action: An Interactive Benchmark for Vision Reason...

【本日の論文】 1. A Very Big Video Reasoning Suite https://huggingface.co/papers/2602.20159 2. VLANeXt: Recipes for Building Strong VLA Models https://huggingface.co/papers/2602.18532 3. SkillOrchestra: Learning to Route Agents via Skill Transfer https://huggingface.co/papers/2602.19672 4. TOPReward: Token Probabilities as Hidden Zero-Shot Rewards for Robotics https://huggingface.co/papers/2602.19...

【本日の論文】 1. VESPO: Variational Sequence-Level Soft Policy Optimization for Stable Off-Policy LLM Training https://huggingface.co/papers/2602.10693 2. Does Your Reasoning Model Implicitly Know When to Stop Thinking? https://huggingface.co/papers/2602.08354 3. Generated Reality: Human-centric World Simulation using Interactive Video Generation with Hand and Camera Control https://huggingface.c...