Jun 2026
LMs as Task-Specific Knowledge Bases: An Interpretability Analysis
The findings suggest that what the model knows and how it is asked are intertwined in parameter space, undermining the "knowledge base"alogy and carrying implications for the reliability and controllability of factual knowledge in LMs.
Amit Elhelo, Amir Globerson, Mor Geva
· arXiv.org · 0 citations