Semantic Regexes: Auto-Interpreting LLM Features with a Structured Language

Angie Boggust, Donghao Ren, Yannick Assogba, Dominik Moritz, Arvind Satyanarayan, Fred Hohman

Keywords: automated interpretability, LLM features, structured languages

TL;DR: We develop a structured language to describe LLM features, resulting in accurate, concise, and consistent descriptions that inherently describe each feature's level of abstraction and help humans reason about LLMs.

Abstract: Automated interpretability aims to translate large language model (LLM) features into human understandable descriptions. However, natural language feature descriptions are often vague, inconsistent, and require manual relabeling. In response, we introduce *semantic regexes*, structured language descriptions of LLM features. By combining primitives that capture linguistic and semantic patterns with modifiers for contextualization, composition, and quantification, semantic regexes produce precise and expressive feature descriptions. Across quantitative benchmarks and qualitative analyses, semantic regexes match the accuracy of natural language while yielding more concise and consistent feature descriptions. Their inherent structure affords new types of analyses, including quantifying feature complexity across layers, scaling automated interpretability from insights into individual features to model-wide patterns. Finally, in user studies, we find that semantic regexes help people build accurate mental models of LLM features.

Primary Area: interpretability and explainable AI

Submission Number: 4063