판 이력 — Decoupling Internal Representational Changes and Causal Importance in Fine-Tuned Large Language Models | AIChainDay