Timezone: »

 
Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small
Kevin Wang · Alexandre Variengien · Arthur Conmy · Buck Shlegeris · Jacob Steinhardt

@
in Workshop on Machine Learning Safety »

Research in mechanistic interpretability seeks to explain behaviors of ML models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task that requires logical reasoning: indirect object identification (IOI). Our explanation encompasses 28 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches including causal interventions and projections. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior ``in the wild" in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks.

[ Poster]  [ Topia] 

Author Information

Kevin Wang
Alexandre Variengien (Redwood Research)
Arthur Conmy (Redwood Research)
Buck Shlegeris (Redwood Research)
Jacob Steinhardt (UC Berkeley)

More from the Same Authors