CausalGym: Benchmarking causal interpretability methods on linguistic tasks

arXiv:2402.12560v1 Announce Type: new
Abstract: Language models (LMs) have proven to be powerful tools for psycholinguistic research, but most prior work has focused on purely behavioural measures (e.g., surprisal comparisons). At the same time, research in model interpretability has begun to illuminate the abstract causal mechanisms shaping LM behavior. To help bring these strands of research closer together, we introduce CausalGym. We adapt and expand the SyntaxGym suite of tasks to benchmark the ability of interpretability methods to causally affect …

abstract arxiv begun behavior benchmarking cs.ai cs.cl interpretability language language models lms model interpretability prior research tasks tools type work