可读规则不该学进LLM权重

重点关注

• AlphaGRPO: Unlocking Self-Reflective Multimodal Generation in UMMs via Decompositional Verifiable Reward score 10
  入选 HF Daily Papers；HF 热度: 32 upvotes (+4)；有代码实现；关键词(3): GRPO, reasoning, text-to-image
• ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents score 10
  入选 HF Daily Papers；HF 热度: 24 upvotes (+4)；有代码实现；关键词(3): scaling, agentic, tool use
• Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling score 10
  入选 HF Daily Papers；HF 热度: 30 upvotes (+4)；有代码实现；关键词(1): reasoning
• L2P: Unlocking Latent Potential for Pixel Generation score 9
  入选 HF Daily Papers；HF 热度: 26 upvotes (+4)；有代码实现
• On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment score 9
  入选 HF Daily Papers；HF 热度: 15 upvotes (+3)；有代码实现；关键词(1): agentic
• Multi-Stream LLMs: Unblocking Language Models with Parallel Streams of Thoughts, Inputs and Outputs score 9
  入选 HF Daily Papers；HF 热度: 16 upvotes (+3)；有代码实现；关键词(1): coding
• Missing Old Logits in Asynchronous Agentic RL: Semantic Mismatch and Repair Methods for Off-Policy Correction score 9
  入选 HF Daily Papers；HF 热度: 15 upvotes (+3)；有代码实现；关键词(3): throughput, PPO, agentic
• Covering Human Action Space for Computer Use: Data Synthesis and Benchmark score 9
  入选 HF Daily Papers；HF 热度: 13 upvotes (+3)；有代码实现；关键词(1): open-source
• Do Enterprise Systems Need Learned World Models? The Importance of Context to Infer Dynamics score 8
  入选 HF Daily Papers；HF 热度: 58 upvotes (+4)；关键词(2): deployment, reasoning
• LoopUS: Recasting Pretrained LLMs into Looped Latent Refinement Models score 8
  入选 HF Daily Papers；HF 热度: 9 upvotes (+2)；有代码实现；关键词(3): scaling, post-training, reasoning

也值得关注

• VidSplat: Gaussian Splatting Reconstruction with Geometry-Guided Video Diffusion Priors score 4
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)
• The DAWN of World-Action Interactive Models score 8
  入选 HF Daily Papers；HF 热度: 19 upvotes (+3)；有代码实现
• MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning score 6
  入选 HF Daily Papers；HF 热度: 7 upvotes (+2)；关键词(1): reasoning
• Context Training with Active Information Seeking score 6
  入选 HF Daily Papers；HF 热度: 5 upvotes (+2)；关键词(2): deployment, reasoning
• Pareto-Guided Optimal Transport for Multi-Reward Alignment score 4
  关键词(1): text-to-image；顶会接收: ICML
• A$_3$B$_2$: Adaptive Asymmetric Adapter for Alleviating Branch Bias in Vision-Language Image Classification with Few-Shot Learning score 4
  关键词(3): lightweight, fine-tuning, vision-language；顶会接收: IJCAI
• Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models score 4
  关键词(2): latency, reasoning；顶会接收: ICML
• PreFIQs: Face Image Quality Is What Survives Pruning score 4
  关键词(1): pruning；顶会接收: CVPR
• RoboEvolve: Co-Evolving Planner-Simulator for Robotic Manipulation with Limited Data score 6
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；关键词(1): vision-language
• Realtime-VLA FLASH: Speculative Inference Framework for Diffusion-based VLAs score 4
  机构: Huawei；关键词(6): lightweight, deployment, latency, real-time, vision-language