TRM Defines Rubrics; SAKI Reaches 4.22× Throughput

Featured

• Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographies score 9
  入选 HF Daily Papers; HF 热度: 76 upvotes (+4); 有代码实现
• Think Before You Score: Thinking Reward Model for Visual Generation score 10
  入选 HF Daily Papers; HF 热度: 86 upvotes (+4); 有代码实现; 关键词(2): GRPO, open-source
• Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVR score 8
  入选 HF Daily Papers; HF 热度: 46 upvotes (+4); 关键词(2): GRPO, reasoning
• OmniTaskonomy: When Does Visual Generation Improve Visual Understanding? score 10
  入选 HF Daily Papers; HF 热度: 45 upvotes (+4); 有代码实现; 关键词(1): reasoning
• SAKI: Maximal-Coupling-Routed Teacher Supervision for On-Policy Distillation score 10
  入选 HF Daily Papers; HF 热度: 82 upvotes (+4); 有代码实现; 关键词(3): distillation, throughput, reasoning
• HiRAE: Hierarchical Representation Autoencoding with Residual Budgets score 7
  入选 HF Daily Papers; HF 热度: 17 upvotes (+3); 关键词(2): fine-tuning, text-to-image
• Real2Gym: Building Gyms from Videos, Bringing Skills to Robots score 6
  入选 HF Daily Papers; 有代码实现; 关键词(1): agentic
• Where the Model Changes Its Mind: Hindsight-Divergence Localization for Efficient Reinforcement Learning with Verifiable Rewards score 5
  入选 HF Daily Papers; HF 热度: 3 upvotes (+1); 关键词(1): GRPO

Also Worth Noting

• Act First, Reason Later: Accelerating On-Policy Distillation for Multi-Turn Agents via Reference-Conditioned Inverse Dynamics score 4
  入选 HF Daily Papers; 关键词(2): distillation, reasoning
• LongLive-Plug: Once-for-All Distillation for Video Generation score 10
  入选 HF Daily Papers; HF 热度: 27 upvotes (+4); 有代码实现; 关键词(3): distillation, deployment, robotics
• Exemplar2VQA: A Scalable Exemplar-Driven Visual Question Answering Generation Framework via Multi-Agent Coding score 4
  关键词(4): fine-tuning, coding, reasoning, embodied; 顶会接收: NeurIPS