10.6k轨迹SFT追平RL流水线

重点关注

• OpenSeeker-v2: Pushing the Limits of Search Agents with Informative and High-Difficulty Trajectories score 10
  入选 HF Daily Papers；HF 热度: 61 upvotes (+4)；有代码实现；关键词(4): scaling, fine-tuning, pre-training, open-source
• PatRe: A Full-Stage Office Action and Rebuttal Generation Benchmark for Patent Examination score 8
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；有代码实现；关键词(2): reasoning, open-source
• StateSMix: Online Lossless Compression via Mamba State Space Models and Sparse N-gram Context Mixing score 8
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；有代码实现；关键词(5): scaling, compression, state space, mamba, coding
• Healthcare AI GYM for Medical Agents score 7
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)；有代码实现；关键词(5): distillation, GRPO, agentic, tool use, reasoning
• The TTS-STT Flywheel: Synthetic Entity-Dense Audio Closes the Indic ASR Gap Where Commercial and Open-Source Systems Fail score 7
  入选 HF Daily Papers；HF 热度: 2 upvotes (+1)；有代码实现；关键词(2): fine-tune, open-source
• SymptomAI: Towards a Conversational AI Agent for Everyday Symptom Assessment score 7
  入选 HF Daily Papers；HF 热度: 10 upvotes (+3)；关键词(1): agentic
• Workspace-Bench 1.0: Benchmarking AI Agents on Workspace Tasks with Large-Scale File Dependencies score 6
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；关键词(1): reasoning

也值得关注

• Generate, Filter, Control, Replay: A Comprehensive Survey of Rollout Strategies for LLM Reinforcement Learning score 6
  入选 HF Daily Papers；HF 热度: 7 upvotes (+2)；关键词(4): throughput, post-training, agentic, reasoning
• iWorld-Bench: A Benchmark for Interactive World Models with a Unified Action Generation Framework score 8
  入选 HF Daily Papers；HF 热度: 4 upvotes (+1)；关键词(2): reasoning, leaderboard；顶会接收: ICML
• AniMatrix: An Anime Video Generation Model that Thinks in Art, Not Physics score 4
  机构: Tencent；关键词(1): production
• Large-Scale High-Quality 3D Gaussian Head Reconstruction from Multi-View Captures score 4
  机构: Apple；关键词(1): scaling