Do Transformers Learn What Theory Predicts? Knowledge Representation-Guided Mechanistic Verification via Causal Abstraction.

Chang Lu, Renate A. Schmidt, Yizheng Zhao

Browse the full KR paper archive.