Sharing Matters: Analysing Neurons Across Languages and Tasks in LLMs

TL;DR

This paper investigates how neurons in multilingual large language models are shared across different languages and tasks, revealing their crucial role in model performance and variability in activation patterns.

Contribution

It introduces a neuron categorization scheme for multilingual LLMs and provides extensive analysis of neuron sharing across languages and tasks, filling a key research gap.

Findings

Deactivating all-shared neurons reduces performance.

Shared neurons are vital for response generation.

Neuron activation varies across tasks, models, and languages.

Abstract

Large language models (LLMs) have revolutionized the field of natural language processing (NLP), and recent studies have aimed to understand their underlying mechanisms. However, most of this research is conducted within a monolingual setting, primarily focusing on English. Few studies have attempted to explore the internal workings of LLMs in multilingual settings. In this study, we aim to fill this research gap by examining how neuron activation is shared across tasks and languages. We classify neurons into four distinct categories based on their responses to a specific input across different languages: all-shared, partial-shared, specific, and non-activated. Building upon this categorisation, we conduct extensive experiments on three tasks across nine languages using several LLMs and present an in-depth analysis in this work. Our findings reveal that: (i) deactivating the all-shared neurons significantly decreases performance; (ii) the shared neurons play a vital role in generating responses, especially for the all-shared neurons; (iii) neuron activation patterns are highly sensitive and vary across tasks, LLMs, and languages. These findings shed light on the internal workings of multilingual LLMs and pave the way for future research. We release the code to foster research in this area.