先写代码再画图，排版提升68%

重点关注

• CARE-Edit: Condition-Aware Routing of Experts for Contextual Image Editing score 13
  入选 HF Daily Papers；HF 热度: 33 upvotes (+4)；有代码实现；关键词(1): lightweight；顶会接收: CVPR
• CoCo: Code as CoT for Text-to-Image Preview and Rare Concept Generation score 10
  入选 HF Daily Papers；HF 热度: 31 upvotes (+4)；有代码实现；关键词(2): reasoning, text-to-image
• HiAR: Efficient Autoregressive Long Video Generation via Hierarchical Denoising score 10
  入选 HF Daily Papers；HF 热度: 26 upvotes (+4)；有代码实现；关键词(1): distillation
• \$OneMillion-Bench: How Far are Language Agents from Human Experts? score 10
  入选 HF Daily Papers；HF 热度: 24 upvotes (+4)；有代码实现；关键词(3): agentic, tool use, reasoning
• Agentic Critical Training score 10
  机构: Zhejiang University；入选 HF Daily Papers；HF 热度: 13 upvotes (+3)；关键词(4): distillation, post-training, agentic, reasoning
• Concept-Guided Fine-Tuning: Steering ViTs away from Spurious Correlations to Improve Robustness score 9
  入选 HF Daily Papers；HF 热度: 11 upvotes (+3)；有代码实现；关键词(3): fine-tuning, finetuning, reasoning
• Scale Space Diffusion score 9
  入选 HF Daily Papers；HF 热度: 10 upvotes (+3)；有代码实现；关键词(1): scaling
• TALON: Test-time Adaptive Learning for On-the-Fly Category Discovery score 9
  入选 HF Daily Papers；有代码实现；关键词(1): quantization；顶会接收: CVPR
• How Far Can Unsupervised RLVR Scale LLM Training? score 8
  入选 HF Daily Papers；HF 热度: 43 upvotes (+4)；关键词(2): scaling, serving
• VLM-SubtleBench: How Far Are VLMs from Human-Level Subtle Comparative Reasoning? score 8
  入选 HF Daily Papers；HF 热度: 9 upvotes (+2)；有代码实现；关键词(3): reasoning, vision-language, open-source

也值得关注

• Beyond Heuristic Prompting: A Concept-Guided Bayesian Framework for Zero-Shot Image Recognition score 4
  关键词(1): vision-language；顶会接收: CVPR
• VisualAD: Language-Free Zero-Shot Anomaly Detection via Vision Transformer score 4
  关键词(2): lightweight, vision-language；顶会接收: CVPR
• It's Time to Get It Right: Improving Analog Clock Reading and Clock-Hand Spatial Reasoning in Vision-Language Models score 4
  关键词(4): fine-tuning, DPO, reasoning, vision-language；顶会接收: CVPR
• WaDi: Weight Direction-aware Distillation for One-step Image Synthesis score 4
  关键词(2): distillation, deployment；顶会接收: CVPR
• Prototype-Guided Concept Erasure in Diffusion Models score 4
  关键词(1): text-to-image；顶会接收: CVPR
• Revealing Behavioral Plasticity in Large Language Models: A Token-Conditional Perspective score 4
  机构: Alibaba；关键词(1): reasoning
• All Vehicles Can Lie: Efficient Adversarial Defense in Fully Untrusted-Vehicle Collaborative Perception via Pseudo-Random Bayesian Inference score 4
  关键词(1): real-time；顶会接收: CVPR
• Impermanent: A Live Benchmark for Temporal Generalization in Time Series Forecasting score 4
  机构: Carnegie Mellon；关键词(1): open-source
• SVG-EAR: Parameter-Free Linear Compensation for Sparse Video Generation via Error-aware Routing score 4
  机构: Tsinghua；关键词(2): lightweight, throughput
• Revisiting Unknowns: Towards Effective and Efficient Open-Set Active Learning score 3
  顶会接收: CVPR
• ELLMob: Event-Driven Human Mobility Generation with Self-Aligned LLM Framework score 3
  顶会接收: ICLR
• Missing No More: Dictionary-Guided Cross-Modal Image Fusion under Missing Infrared score 3
  顶会接收: CVPR
• Model-based Offline RL via Robust Value-Aware Model Learning with Implicitly Differentiable Adaptive Weighting score 3
  顶会接收: ICLR
• Gender Bias in MT for a Genderless Language: New Benchmarks for Basque score 3
  机构: FAIR
• SiMO: Single-Modality-Operable Multimodal Collaborative Perception score 3
  顶会接收: ICLR