WorldTrace让场景召回提升19.5%

重点关注

• Modular TTT: Rethinking Test-Time Training as Composable Modules score 7
  机构: ByteDance；入选 HF Daily Papers；HF 热度: 2 upvotes (+1)
• Addressable Memory for Video World Models score 7
  机构: NVIDIA；入选 HF Daily Papers；关键词(1): compression
• The Optimizer Is the Agent: Reasoning-Driven Search across Prompts, Programs, and ML Workflows score 5
  入选 HF Daily Papers；HF 热度: 2 upvotes (+1)；关键词(2): serving, reasoning

也值得关注

• Skaling: Chinchilla's Exponents Meet Kaplan's Coupling score 4
  入选 HF Daily Papers；关键词(1): scaling
• Bend the Basics: Degradation-Aware Deformable Tokenization for All-in-One Image Restoration score 4
  关键词(1): lightweight；顶会接收: ICML
• How Should I Pick a Foundation Model for My Robot? In Favor of a Community Evaluation Framework for Social Robots score 4
  机构: Google；关键词(2): real-time, embodied
• Prune Once: Retraining-Free Task-Agnostic Pruning for Vision-Language Models score 4
  关键词(5): compression, pruning, deployment, pre-training, vision-language；顶会接收: ECCV
• HRDiT: Training-Free High-Resolution Image Generation with Off-the-Shelf Diffusion Transformer Models score 4
  关键词(1): text-to-image；顶会接收: ECCV
• Suppress and Diversify: Refining Robust Pathways for Corruption Robustness score 3
  顶会接收: ICML
• Autonomous discovery of accelerator commissioning algorithms score 3
  机构: Berkeley