Model interpretability and model editing are crucial goals in the age of large language models. Interestingly, there exists a link between these two goals: if a method is able to systematically edit model behavior with regard to a human concept of interest, this editor method can help make internal representations more interpretable by pointing towards relevant representations and systematically manipulating them.
@article{arxiv.2311.10905,
title = {Flexible Model Interpretability through Natural Language Model Editing},
author = {Karel D'Oosterlinck and Thomas Demeester and Chris Develder and Christopher Potts},
journal= {arXiv preprint arXiv:2311.10905},
year = {2023}
}
Comments
Extended Abstract -- work in progress. BlackboxNLP2023