递归MAS省35%token，T2I整张重画

重点关注

• Recursive Multi-Agent Systems score 10
  入选 HF Daily Papers；HF 热度: 200 upvotes (+4)；有代码实现；关键词(4): scaling, lightweight, code generation, reasoning
• Refinement via Regeneration: Enlarging Modification Space Boosts Image Refinement in Unified Multimodal Models score 10
  入选 HF Daily Papers；HF 热度: 22 upvotes (+4)；有代码实现；关键词(1): text-to-image
• Mutual Forcing: Dual-Mode Self-Evolution for Fast Autoregressive Audio-Video Character Generation score 9
  入选 HF Daily Papers；HF 热度: 15 upvotes (+3)；有代码实现；关键词(1): distillation
• BARRED: Synthetic Training of Custom Policy Guardrails via Asymmetric Debate score 8
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；有代码实现；关键词(4): fine-tuning, reasoning, guardrails, synthetic data
• DV-World: Benchmarking Data Visualization Agents in Real-World Scenarios score 9
  入选 HF Daily Papers；HF 热度: 37 upvotes (+4)；有代码实现

也值得关注

• Toward Scalable Terminal Task Synthesis via Skill Graphs score 6
  入选 HF Daily Papers；HF 热度: 7 upvotes (+2)；关键词(2): scaling, agentic
• FAMA: Failure-Aware Meta-Agentic Framework for Open-Source LLMs in Interactive Tool Use Environments score 9
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；关键词(3): agentic, tool use, open-source；顶会接收: ACL
• Do LLMs Capture Embodied Cognition and Cultural Variation? Cross-Linguistic Evidence from Demonstratives score 4
  关键词(2): reasoning, embodied；顶会接收: ACL
• A Systematic Post-Train Framework for Video Generation score 4
  入选 HF Daily Papers；关键词(6): deployment, fine-tuning, RLHF, GRPO, post-training
• Prefill-Time Intervention for Mitigating Hallucination in Large Vision-Language Models score 4
  关键词(1): vision-language；顶会接收: CVPR
• CORAL: Adaptive Retrieval Loop for Culturally-Aligned Multilingual RAG score 4
  关键词(3): retrieval-augmented, RAG, agentic；顶会接收: ACL
• When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient score 4
  机构: Princeton；关键词(1): RLHF
• IAM: Identity-Aware Human Motion and Shape Joint Generation score 3
  入选 HF Daily Papers