模型压缩的几何直觉：折叠比剪枝更优

重点关注

• Generated Reality: Human-centric World Simulation using Interactive Video Generation with Hand and Camera Control score 8
  入选 HF Daily Papers；HF 热度: 17 upvotes (+3)；关键词(3): transformer, diffusion, embodied
• Learning Optimal and Sample-Efficient Decision Policies with Guarantees score 5
  机构: Oxford；关键词(5): efficient, efficiency, agent, robotics, evaluation
• Interacting safely with cyclists using Hamilton-Jacobi reachability and reinforcement learning score 5
  机构: Carnegie Mellon；关键词(2): efficient, safety
• Cut Less, Fold More: Model Compression through the Lens of Projection Geometry score 5
  关键词(3): compression, pruning, deployment；顶会接收: ICLR
• DEIG: Detail-Enhanced Instance Generation with Fine-Grained Semantic Control score 5
  关键词(3): attention, diffusion, benchmark；顶会接收: AAAI
• SARAH: Spatially Aware Real-time Agentic Humans score 6
  入选 HF Daily Papers；HF 热度: 4 upvotes (+1)；关键词(11): deployment, inference, real-time, alignment, transformer

也值得关注

• Flow Actor-Critic for Offline Reinforcement Learning score 3
  顶会接收: ICLR
• A Single Image and Multimodality Is All You Need for Novel View Synthesis score 2
  关键词(4): efficient, inference, diffusion, multimodal
• Alignment in Time: Peak-Aware Orchestration for Long-Horizon Agentic Systems score 2
  关键词(5): alignment, agent, agents, agentic, evaluation
• Condition-Gated Reasoning for Context-Dependent Biomedical Question Answering score 2
  关键词(3): retrieval-augmented, reasoning, benchmark
• From Lossy to Verified: A Provenance-Aware Tiered Memory for Agents score 2
  关键词(6): fast, compression, inference, latency, agent
• Efficient Filtered-ANN via Learning-based Query Planning score 2
  关键词(4): efficient, lightweight, planning, search
• Latent Diffeomorphic Co-Design of End-Effectors for Deformable and Fragile Object Manipulation score 2
  关键词(3): distillation, deployment, robotics
• ZACH-ViT: Regime-Dependent Inductive Bias in Compact Vision Transformers for Medical Imaging score 2
  关键词(7): deployment, inference, edge, pretraining, transformer
• MIRA: Memory-Integrated Reinforcement Learning Agent with Limited LLM Guidance score 2
  关键词(3): real-time, agent, agents
• Memory-Based Advantage Shaping for LLM-Guided Reinforcement Learning score 2
  关键词(3): efficiency, agent, benchmark
• Causal Neighbourhood Learning for Invariant Graph Representations score 2
  关键词(2): edge, attention
• Analyzing LLM Instruction Optimization for Tabular Fact Verification score 2
  关键词(3): lightweight, agents, reasoning
• CUICurate: A GraphRAG-based Framework for Automated Clinical Concept Curation for NLP applications score 2
  关键词(3): retrieval-augmented, reasoning, benchmark
• ROCKET: Residual-Oriented Multi-Layer Alignment for Spatially-Aware Vision-Language-Action Models score 2
  关键词(2): alignment, vision-language
• Hardware-Friendly Input Expansion for Accelerating Function Approximation score 2
  关键词(2): efficient, cost