#artificial intelligence
Jul 2025
Can Interpretation Predict Behavior on Unseen Data?
This work proposes and demonstrates an alternate objective for a new interpretability objective: understanding model internals to predict behavior and assess reliability under distribution shift, and successfully uses attention patterns to predict which rule each model follows on OOD data.
Victoria R. Li, Jenny Kaufmann, Tian Qin et al.
· arXiv.org · 13 citations
· ⚡1