Can LLMs Reconcile Knowledge Conflicts in Counterfactual Reasoning

Khurram Yamin, Gaurav Ghosal, Bryan Wilder

Large Language Models have been shown to contain extensive world knowledge in their parameters, enabling impressive performance on many knowledge intensive tasks. However, when dep…