Mechanistic Interpretability

CRAFT: Causal Responsibility and Failure Tracing in Medical Vision Language Models
Causal tracing and targeted intervention for arbitration and brake failures in medical vision language models.
Chunzheng Zhu, Jiaqi Zeng, Hongbo Zhao, Yihang Chen, Yijun Wang, Jianxin Lin
CRAFT: Causal Responsibility and Failure Tracing in Medical Vision Language Models