SpecMRL半维embedding胜全维普通MRL

重点关注

• MemT-1.0: Towards Memory-Native Evolution score 13
  机构: Zhejiang University；入选 HF Daily Papers；HF 热度: 131 upvotes (+4)；有代码实现；关键词(4): post-training, pre-training, retrieval-augmented, agentic
• HarnessPro: Recursive Harness Self-Improvement with Pareto Optimization for Performance-Cost Trade-offs score 13
  机构: Zhipu；入选 HF Daily Papers；HF 热度: 30 upvotes (+4)；有代码实现；关键词(1): deployment
• BlenderWorld: Benchmarking Visual Agents for 3D Creation from Real-World References score 12
  机构: Shanghai AI Lab；入选 HF Daily Papers；HF 热度: 35 upvotes (+4)；有代码实现
• OR-License: Evaluating LLM Agents as Real-World Operations Research Practitioners score 10
  入选 HF Daily Papers；HF 热度: 292 upvotes (+4)；有代码实现；关键词(2): deployment, coding
• OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory score 10
  入选 HF Daily Papers；HF 热度: 110 upvotes (+4)；有代码实现；关键词(3): agentic, coding, reasoning
• SpecMRL: Redesign Matryoshka Representation Learning within Parametric Spectrum score 10
  入选 HF Daily Papers；HF 热度: 85 upvotes (+4)；有代码实现；关键词(1): compression
• AnyGoal: On World Model Generalization Beyond Expert Rollouts score 10
  入选 HF Daily Papers；HF 热度: 42 upvotes (+4)；有代码实现；关键词(1): open-source
• AdaExploreBench: Benchmarking LLM Agents for Proactive Exploration and Test-time Adaptation in Evolving Environments score 9
  入选 HF Daily Papers；HF 热度: 33 upvotes (+4)；有代码实现
• Remembering Has a Price: Learning Memory Worth under Resource Constraints score 9
  入选 HF Daily Papers；HF 热度: 11 upvotes (+3)；有代码实现；关键词(3): lightweight, deployment, reasoning
• Text-to-Image Generative Modeling Without Language Pretraining score 9
  机构: Stanford；入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；关键词(2): pretraining, text-to-image

也值得关注

• World of SWECraft: Agent Collaboration at Scale via Simulated Software Worlds score 4
  机构: Carnegie Mellon；关键词(1): scaling
• BLANC: Adaptive Temporal Scheduling for Sign Language Production score 4
  机构: Carnegie Mellon；关键词(1): production
• CoDA: A Human-LM Collaborative Codebook Development Assistant for Domain-Specific Text Annotation score 4
  机构: Allen Institute；关键词(1): open-source
• Self-Improving Video-to-Action Models score 4
  机构: MIT；关键词(1): embodied
• Learning to Reason Efficiently with Interventional Step Credit score 4
  机构: Cambridge；关键词(1): reasoning
• Maestro: Agentic Formal Proof Development by Coordination score 4
  机构: Amazon；关键词(2): agentic, coding
• Scaling Synthetic Pre-training Data and Evaluation for Indic LLMs score 4
  机构: Microsoft；关键词(4): scaling, pre-training, pretraining, reasoning
• Physiology-Aware Graph Attention for Early Ventricular Reentry Prediction score 4
  机构: INRIA；关键词(1): edge
• When the harness decides success: state that cannot be re-observed score 4
  机构: Microsoft；关键词(1): reasoning
• LatComm: Reusable Context Compression for Latent Communication between Large Language Models score 4
  机构: Huawei；关键词(1): compression
• TRACE: Tracking Multimodal Clinical Trajectories for Inpatient Risk Prediction score 3
  入选 HF Daily Papers
• Forecasting Patient Trajectories from the Electronic Health Records with Action-Conditioned World Models score 3
  机构: Google
• HumanSkill-4D: Training-Free Language Querying of Animatable Gaussian Avatars in Space and Time score 4
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)
• Evidence Sufficiency and Model Comparisons in MexRegQA: An Item-Level Audit of Mexican-Spanish Regulatory QA score 3
  机构: Huawei