Circuit Component Reuse Across Tasks in Transformer Language Models

arXiv:2310.08744v3 Announce Type: replace-cross
Abstract: Recent work in mechanistic interpretability has shown that behaviors in language models can be successfully reverse-engineered through circuit analysis. A common criticism, however, is that each circuit is task-specific, and thus such analysis cannot contribute to understanding the models at a higher level. In this work, we present evidence that insights (both low-level findings about specific heads and higher-level findings about general algorithms) can indeed generalize across tasks. Specifically, we study the circuit discovered in …

abstract analysis arxiv cs.cl cs.lg however interpretability language language models tasks through transformer transformer language models type understanding work