AI论文速递 2026年06月02日（HuggingFace Daily Papers）¶
数据来源：https://huggingface.co/papers 采集时间：2026-06-02
📌 重点关注¶
- SAAS: Self-Aware Reinforcement Learning for Over-Search Mitigation in Agentic Search | arXiv — 【重点关注】 Agentic search enables LLMs to solve complex multi-hop questions through iter... 💡 自搜索强化学习，对Agent系统优化有价值
- Mellum2 Technical Report | arXiv — 【重点关注】 We present Mellum 2, an open-weight 12B-parameter Mixture-of-Experts (MoE) la... 💡 12B MoE开源模型，实用性强的架构选择
- Harness Updating Is Not Harness Benefit: Disentangling Evolution Capabilities in Self-Evolving LLM Agents | arXiv — 【重点关注】 LLM agents are increasingly deployed as systems built around editable externa... 💡 重新思考Agent进化能力，理论突破
📋 其他值得关注¶
- VisualThink-VLA: Visual Intermediate Reasoning for Effective and Low-Latency Vision-Language-Action Policies | arXiv — Recent work has begun to equip vision-language-action (VLA) policies with exp...
- GrepSeek: Training Search Agents for Direct Corpus Interaction | arXiv — Large Language Model (LLM) search agents have shown strong promise for knowle...
- SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories | arXiv — Large language model (LLM) agents increasingly rely on reusable external skil...
- RoboStressBench: Benchmarking VLM Robustness to Physical Visual Stress in Embodied Scenes | arXiv — Vision-Language Models (VLMs) have shown strong visual understanding and are ...
- From Model Scaling to System Scaling: Scaling the Harness in Agentic AI | arXiv — This paper studies the next major bottleneck in agentic AI as system scaling,...
- Memory-Bound but Not Bandwidth-Limited: The Physical AI Inference Gap in Batch-1 LLM Decode | arXiv — Physical AI systems, including robots, autonomous vehicles, embodied agents a...
- Flat-Pack Bench: Evaluating Spatio-Temporal Understanding in Large Vision-Language Models through Furniture Assembly | arXiv — The emergence of Large Vision-Language Models (LVLMs) has significantly advan...