用扩散模型做投机解码，推理加速6倍

重点关注

• Spider-Sense: Intrinsic Risk Sensing for Efficient Agent Defense with Hierarchical Adaptive Screening score 9
  入选 HF Daily Papers；HF 热度: 63 upvotes (+4)；关键词(9): efficient, efficiency, lightweight, latency, agent
• Length-Unbiased Sequence Policy Optimization: Revealing and Controlling Response Length Variation in RLVR score 9
  入选 HF Daily Papers；HF 热度: 45 upvotes (+4)；关键词(4): GRPO, reasoning, multimodal, vision-language
• Context Forcing: Consistent Autoregressive Video Generation with Long Context score 9
  入选 HF Daily Papers；HF 热度: 29 upvotes (+4)；关键词(3): fast, real-time, evaluation
• DFlash: Block Diffusion for Flash Speculative Decoding score 9
  入选 HF Daily Papers；HF 热度: 27 upvotes (+4)；关键词(6): efficient, fast, lightweight, inference, latency
• RISE-Video: Can Video Generators Decode Implicit World Rules? score 9
  入选 HF Daily Papers；HF 热度: 26 upvotes (+4)；关键词(5): alignment, reasoning, multimodal, benchmark, evaluation
• ProAct: Agentic Lookahead in Interactive Environments score 9
  入选 HF Daily Papers；HF 热度: 22 upvotes (+4)；关键词(13): lightweight, distillation, inference, fine-tuning, PPO
• Dr. Kernel: Reinforcement Learning Done Right for Triton Kernel Generations score 9
  入选 HF Daily Papers；HF 热度: 20 upvotes (+4)；关键词(2): scaling, GRPO
• InterPrior: Scaling Generative Control for Physics-Based Human-Object Interactions score 8
  入选 HF Daily Papers；HF 热度: 18 upvotes (+3)；关键词(6): scaling, deployment, finetuning, post-training, pretraining
• Semantic Search over 9 Million Mathematical Theorems score 8
  入选 HF Daily Papers；HF 热度: 17 upvotes (+3)；关键词(3): agents, search, evaluation
• Reinforcement World Model Learning for LLM-based Agents score 8
  入选 HF Daily Papers；HF 热度: 17 upvotes (+3)；关键词(2): agents, agentic

也值得关注

• AI chatbots versus human healthcare professionals: a systematic review and meta-analysis of empathy in patient care score 3
  机构: Meta
• Verification of the Implicit World Model in a Generative Model via Adversarial Sequences score 3
  顶会接收: ICLR
• TIDE: Temporal Incremental Draft Engine for Self-Improving LLM Inference score 2
  关键词(3): serving, inference, throughput
• CoSA: Compressed Sensing-Based Adaptation of Large Language Models score 2
  关键词(2): efficient, fine-tuning
• GreekMMLU: A Native-Sourced Multitask Benchmark for Evaluating Language Models in Greek score 2
  关键词(3): benchmark, evaluation, leaderboard
• RAG without Forgetting: Continual Query-Infused Key Memory score 2
  关键词(5): inference, retrieval-augmented, RAG, reasoning, cost
• AirGlove: Exploring Egocentric 3D Hand Tracking and Appearance Generalization for Sensing Gloves score 2
  关键词(3): fine-tuning, pre-training, evaluation
• LOBSTgER-enhance: an underwater image enhancement pipeline score 2
  关键词(2): diffusion, evaluation
• Position: Capability Control Should be a Separate Goal From Alignment score 2
  关键词(4): deployment, alignment, preference, guardrails
• EBPO: Empirical Bayes Shrinkage for Stabilizing Group-Relative Policy Optimization score 2
  关键词(2): GRPO, reasoning
• ShapePuri: Shape Guided and Appearance Generalized Adversarial Purification score 2
  关键词(5): efficient, inference, diffusion, benchmark, cost
• Among Us: Measuring and Mitigating Malicious Contributions in Model Collaboration Systems score 2
  关键词(3): agent, reasoning, safety