Flow-OPD把GenEval从63拉到92

重点关注

• Flow-OPD: On-Policy Distillation for Flow Matching Models score 10
  入选 HF Daily Papers；HF 热度: 88 upvotes (+4)；有代码实现；关键词(5): distillation, fine-tuning, GRPO, post-training, text-to-image
• HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents score 10
  入选 HF Daily Papers；HF 热度: 57 upvotes (+4)；有代码实现；关键词(2): distillation, open-source
• LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling score 10
  入选 HF Daily Papers；HF 热度: 62 upvotes (+4)；有代码实现；关键词(4): scaling, agentic, reasoning, open-source
• HumanNet: Scaling Human-centric Video Learning to One Million Hours score 10
  入选 HF Daily Papers；HF 热度: 48 upvotes (+4)；有代码实现；关键词(5): scaling, tool use, vision-language, embodied, data curation
• A$^2$RD: Agentic Autoregressive Diffusion for Long Video Consistency score 10
  机构: Yale；入选 HF Daily Papers；HF 热度: 15 upvotes (+3)；关键词(1): agentic
• MatryoshkaLoRA: Learning Accurate Hierarchical Low-Rank Representations for LLM Fine-Tuning score 9
  入选 HF Daily Papers；HF 热度: 17 upvotes (+3)；有代码实现；关键词(2): deployment, fine-tuning
• SCOPE: Structured Decomposition and Conditional Skill Orchestration for Complex Image Generation score 8
  入选 HF Daily Papers；HF 热度: 9 upvotes (+2)；有代码实现；关键词(2): reasoning, text-to-image
• IntentGrasp: A Comprehensive Benchmark for Intent Understanding score 8
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；有代码实现；关键词(1): fine-tuning
• ModelLens: Finding the Best for Your Task from Myriads of Models score 8
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；有代码实现；关键词(3): vision-language, leaderboard, open-source
• InterLV-Search: Benchmarking Interleaved Multimodal Agentic Search score 8
  入选 HF Daily Papers；HF 热度: 5 upvotes (+2)；有代码实现；关键词(3): agentic, tool use, open-source

也值得关注

• BalCapRL: A Balanced Framework for RL-Based MLLM Image Captioning score 5
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)；关键词(1): GRPO
• Tool Calling is Linearly Readable and Steerable in Language Models score 4
  机构: Imperial College；关键词(3): instruction tuning, pretraining, agentic
• PACEvolve++: Improving Test-time Learning for Evolutionary Search Agents score 4
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)
• AGWM: Affordance-Grounded World Models for Environments with Compositional Prerequisites score 3
  机构: Amazon