验证比发现便宜，4B打平大模型

重点关注

• Show, Don't Tell: Evaluating Spatial Cognition in Generative Pixels Rather Than LLM Text score 10
  入选 HF Daily Papers；HF 热度: 35 upvotes (+4)；有代码实现；关键词(2): agentic, reasoning
• AREX: Towards a Recursively Self-Improving Agent for Deep Research score 10
  入选 HF Daily Papers；HF 热度: 131 upvotes (+4)；有代码实现；关键词(2): agentic, reasoning
• Visual Contrastive Self-Distillation score 10
  入选 HF Daily Papers；HF 热度: 43 upvotes (+4)；有代码实现；关键词(2): distillation, reasoning
• Tencent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction score 11
  机构: Tencent；入选 HF Daily Papers；HF 热度: 20 upvotes (+4)；关键词(2): coding, leaderboard
• LLMs Get Lost in Evolving User Intent score 8
  入选 HF Daily Papers；HF 热度: 19 upvotes (+3)；有代码实现
• Self-Supervised Learning of Structured Dynamics from Videos score 8
  入选 HF Daily Papers；HF 热度: 18 upvotes (+3)；有代码实现
• ICAE-Bench: Evaluating Coding Agents as Interactive Project Builders score 8
  入选 HF Daily Papers；HF 热度: 5 upvotes (+2)；有代码实现；关键词(3): tool use, coding, open-source
• NVIDIA-labs OO Agents: Native Python Object-Oriented Agents score 8
  入选 HF Daily Papers；HF 热度: 25 upvotes (+4)；关键词(2): agentic, reasoning
• SANA-Video 2.0: Hybrid Linear Attention with Attention Residuals for Efficient Video Generation score 8
  入选 HF Daily Papers；HF 热度: 26 upvotes (+4)；关键词(2): scaling, latency
• Recurrent Sinusoidal INRs for Efficient High-Fidelity Representation score 10
  入选 HF Daily Papers；HF 热度: 6 upvotes (+2)；有代码实现；顶会接收: ECCV