12B超GPT-4，蒸馏后学生反超教师

重点关注

• A Subgoal-driven Framework for Improving Long-Horizon LLM Agents score 5
  入选 HF Daily Papers；HF 热度: 4 upvotes (+1)；关键词(3): real-time, fine-tuning, reasoning
• WorldAgents: Can Foundation Image Models be Agents for 3D World Models? score 5
  入选 HF Daily Papers；HF 热度: 2 upvotes (+1)；关键词(2): agentic, vision-language
• How Well Does Generative Recommendation Generalize? score 5
  入选 HF Daily Papers；有代码实现

也值得关注

• Beyond Single Tokens: Distilling Discrete Diffusion Models via Discrete MMD score 4
  入选 HF Daily Papers；关键词(1): distillation
• LumosX: Relate Any Identities with Their Attributes for Personalized Video Generation score 4
  入选 HF Daily Papers；关键词(1): text-to-video
• FlowScene: Style-Consistent Indoor Scene Generation with Multimodal Graph Rectified Flow score 4
  入选 HF Daily Papers；关键词(1): reasoning
• Semantic Audio-Visual Navigation in Continuous Environments score 4
  关键词(2): reasoning, embodied；顶会接收: CVPR
• Vision-Language Attribute Disentanglement and Reinforcement for Lifelong Person Re-Identification score 4
  关键词(1): vision-language；顶会接收: CVPR
• Adapting a Pre-trained Single-Cell Foundation Model to Spatial Gene Expression Generation from Histology Images score 4
  关键词(3): lightweight, throughput, pre-training；顶会接收: CVPR
• An Agentic Multi-Agent Architecture for Cybersecurity Risk Management score 4
  机构: Microsoft；关键词(1): agentic
• VideoSeek: Long-Horizon Video Agent with Tool-Guided Seeking score 4
  关键词(2): agentic, reasoning；顶会接收: CVPR
• Disentangle-then-Align: Non-Iterative Hybrid Multimodal Image Registration via Cross-Scale Feature Disentanglement score 3
  顶会接收: CVPR
• 3D Gaussian Splatting with Self-Constrained Priors for High Fidelity Surface Reconstruction score 3
  顶会接收: CVPR
• BALM: A Model-Agnostic Framework for Balanced Multimodal Learning under Imbalanced Missing Rates score 3
  顶会接收: CVPR
• PerformRecast: Expression and Head Pose Disentanglement for Portrait Video Editing score 3
  顶会接收: CVPR
• FlashCap: Millisecond-Accurate Human Motion Capture via Flashing LEDs and Event-Based Vision score 3
  顶会接收: CVPR
• Failure Modes for Deep Learning-Based Online Mapping: How to Measure and Address Them score 3
  顶会接收: CVPR
• NEC-Diff: Noise-Robust Event-RAW Complementary Diffusion for Seeing Motion in Extreme Darkness score 3
  顶会接收: CVPR