Lorem Ipsum救回GRPO难题样本

重点关注

• Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction score 10
  入选 HF Daily Papers；HF 热度: 74 upvotes (+4)；有代码实现；关键词(3): lightweight, agentic, reasoning
• MARBLE: Multi-Aspect Reward Balance for Diffusion RL score 10
  入选 HF Daily Papers；HF 热度: 34 upvotes (+4)；有代码实现；关键词(1): fine-tuning
• Continuous-Time Distribution Matching for Few-Step Diffusion Distillation score 10
  入选 HF Daily Papers；HF 热度: 24 upvotes (+4)；有代码实现；关键词(1): distillation
• StraTA: Incentivizing Agentic Reinforcement Learning with Strategic Trajectory Abstraction score 9
  入选 HF Daily Papers；HF 热度: 17 upvotes (+3)；有代码实现；关键词(2): GRPO, agentic
• Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning score 8
  入选 HF Daily Papers；HF 热度: 63 upvotes (+4)；关键词(1): distillation
• Continuous Latent Diffusion Language Model score 8
  入选 HF Daily Papers；HF 热度: 59 upvotes (+4)；关键词(2): scaling, compression
• MiA-Signature: Approximating Global Activation for Long-Context Understanding score 8
  入选 HF Daily Papers；HF 热度: 49 upvotes (+4)；关键词(3): lightweight, RAG, agentic
• SkillOS: Learning Skill Curation for Self-Evolving Agents score 8
  入选 HF Daily Papers；HF 热度: 32 upvotes (+4)；关键词(2): agentic, reasoning
• Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration score 8
  入选 HF Daily Papers；HF 热度: 31 upvotes (+4)；关键词(2): GRPO, reasoning
• Auto Research with Specialist Agents Develops Effective and Non-Trivial Training Recipes score 8
  入选 HF Daily Papers；HF 热度: 12 upvotes (+3)；有代码实现

也值得关注

• Think, then Score: Decoupled Reasoning and Scoring for Video Reward Modeling score 5
  入选 HF Daily Papers；HF 热度: 2 upvotes (+1)；关键词(3): scaling, post-training, reasoning
• TIDE: Every Layer Knows the Token Beneath the Context score 4
  入选 HF Daily Papers；HF 热度: 4 upvotes (+1)