AI.info

Haining Yu

Explore Haining Yu on AI.info.

• Towards Long-Horizon Interpretability: Efficient and Faithful Multi-Token Attribution for Reasoning LLMs