On the Similarity of Circuits across Languages: a Case Study on the Subject-verb Agreement Task
Keywords: circuits, mechanistic interpretability, multilinguality
TL;DR: A study about how a language model solves the subject-verb agreement task in different languages
Abstract: Several algorithms implemented by language models have recently been successfully reversed-engineered. However, these findings have been concentrated on specific tasks and models, leaving it unclear how \textit{universal} circuits are across different settings. In this paper, we study the circuits implemented by Gemma 2B for solving the subject-verb agreement task across two different languages, English and Spanish. We discover that both circuits are highly consistent, being mainly driven by a particular attention head writing a `subject number' signal to the last residual stream, which is read by a small set of neurons in the final MLP layers. Notably, this subject number signal is represented as a direction in the residual stream space, and is language-independent. Finally, we demonstrate this direction has a causal effect on the model predictions, effectively flipping the Spanish predicted verb number by intervening with the direction found in English examples.
Submission Number: 49
Loading