English

TopEx: Topic-based Explanations for Model Comparison

Computation and Language 2023-06-05 v2

Abstract

Meaningfully comparing language models is challenging with current explanation methods. Current explanations are overwhelming for humans due to large vocabularies or incomparable across models. We present TopEx, an explanation method that enables a level playing field for comparing language models via model-agnostic topics. We demonstrate how TopEx can identify similarities and differences between DistilRoBERTa and GPT-2 on a variety of NLP tasks.

Cite

@article{arxiv.2306.00976,
  title  = {TopEx: Topic-based Explanations for Model Comparison},
  author = {Shreya Havaldar and Adam Stein and Eric Wong and Lyle Ungar},
  journal= {arXiv preprint arXiv:2306.00976},
  year   = {2023}
}

Comments

Accepted to ICLR 2023, Tiny Papers Track

R2 v1 2026-06-28T10:53:45.842Z