What Happened in LLMs Layers when Trained for Fast vs. Slow Thinking: A Gradient Perspective.

Ming Li, Yanhong Li, Tianyi Zhou

Browse the full ACL paper archive.