VLM数小时控机器人，世界状态准确率98%

重点关注

• Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation score 10
  入选 HF Daily Papers；HF 热度: 87 upvotes (+4)；有代码实现；关键词(2): distillation, post-training
• SceneMosaic: Efficient and Diverse Simulation-Ready Scene Generation via Hybrid Agentic Layout Evolution score 10
  入选 HF Daily Papers；HF 热度: 42 upvotes (+4)；有代码实现；关键词(3): agentic, vision-language, embodied
• Kalman Delta Networks: Uncertainty-aware Associative Memory score 10
  入选 HF Daily Papers；HF 热度: 27 upvotes (+4)；有代码实现；关键词(1): pretraining
• Agentic Visual Generation: From Generative Models to Agentic Control score 10
  入选 HF Daily Papers；HF 热度: 27 upvotes (+4)；有代码实现；关键词(2): agentic, tool use
• Show-Harness: Just a VLM Agent Can Play Robots score 10
  入选 HF Daily Papers；HF 热度: 134 upvotes (+4)；有代码实现；关键词(7): deployment, fine-tuning, pretraining, agentic, vision-language
• VidaForge: Open Research Infrastructure for Video Pretraining Data Recipes score 9
  入选 HF Daily Papers；HF 热度: 18 upvotes (+3)；有代码实现；关键词(1): pretraining
• TransNormal-2: Geometry-Grounded Rectified Flow with Edge-Aware Decoding for Precise Normal Estimation score 9
  入选 HF Daily Papers；HF 热度: 18 upvotes (+3)；有代码实现；关键词(3): lightweight, compression, edge
• Multi-Grid Post-Training for Long-Form Multi-Shot Video Generation score 9
  入选 HF Daily Papers；HF 热度: 17 upvotes (+3)；有代码实现；关键词(1): post-training
• Programmable World Model score 10
  入选 HF Daily Papers；HF 热度: 85 upvotes (+4)；有代码实现；关键词(2): lightweight, serving
• Online Draft Co-Training for Speculative Decoding in Large-Scale, Long-Context RL Post-Training score 8
  入选 HF Daily Papers；HF 热度: 28 upvotes (+4)；关键词(2): scaling, post-training

也值得关注

• NOAH: Learning the Full Patient Journey. A Longitudinal Multimodal Time-Aware Model for Representation and Forecasting score 4
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)
• BuzzASR: A Swarm of 100+ Monolingual Speech Recognition Models score 4
  关键词(3): compression, fine-tuning, open-source；顶会接收: EMNLP
• PRAGMA: Evaluating Personalized Guidance with Memory Alignment in Lifelong Conversations score 4
  关键词(1): reasoning；顶会接收: EMNLP
• SEA-SpeechBench: A Large-Scale Multitask Benchmark for Speech Understanding Across Southeast Asia score 4
  关键词(1): open-source；顶会接收: EMNLP
• EFQ-Softmax: Exp-Free Quantization for Softmax score 4
  机构: Huawei；关键词(2): quantization, latency
• When Does Defendant Statement Matter? A Study of Bias and Persuasion in LLM-Simulated Jurors score 4
  关键词(1): reasoning；顶会接收: EMNLP
• 5-Dialects-BN: Unmasking the Impact of Transliteration on Bangla Dialectal LLMs score 4
  关键词(1): fine-tuning；顶会接收: EMNLP
• Direct Diversity Optimization for Diverse Successful Trajectories in Preference Post-Training score 4
  关键词(1): post-training；顶会接收: EMNLP
• JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition score 4
  关键词(3): lightweight, reasoning, open-source；顶会接收: EMNLP
• Single Image to Textured 3D Object Generation in Frequency Domain: From Theory to Pipeline score 3
  机构: Peking University
• Compact Visuotactile World Models for Lifting: Prediction, Reward Alignment, and Force Constraints score 3
  机构: Zhejiang University
• Orukeet: Multilingual ASR with Frozen Gabor Kernels score 3
  机构: Stanford
• YallaMorph: A Benchmark for Evaluating Arabic Morphological Generation in Large Language Models score 3
  顶会接收: EMNLP