Sharing Matters: Analysing Neurons Across Languages and Tasks in LLMs

2024-06-13

arXiv.org

AI Summary

This study investigates neuron activation sharing across languages and tasks in multilingual large language models (LLMs), addressing the gap in interpretability research—largely confined to English-only settings. We propose a four-category neuron taxonomy—fully shared, partially shared, language- or task-specific, and inactive—based on activation patterns, and employ clustering analysis, cross-lingual response comparison, and Integrated Gradients attribution to systematically characterize sharing regularities. Contrary to expectations, we find that neuron sharing is independent of linguistic genealogy; fully shared neurons significantly enhance generation quality. Empirical results indicate that task type and input sample characteristics—not language identity—primarily govern activation sharing patterns. This work breaks the monolingual interpretability paradigm, offering novel theoretical insights and methodological pathways for understanding multilingual model mechanisms and enabling efficient, cross-lingual fine-tuning.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

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 attempt to explore the internal workings of LLMs in multilingual settings. In this study, we aim to fill the 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. This categorization is combined with a study of neuron attribution, i.e. the importance of a neuron w.r.t an output. Our analysis reveals the following insights: (i) the patterns of neuron sharing are significantly affected by the characteristics of tasks and examples; (ii) neuron sharing does not fully correspond with language similarity; (iii) shared neurons play a vital role in generating responses, especially those shared across all languages. These findings shed light on the internal workings of multilingual LLMs and pave the way to the future research. We will release the code to foster research in this area.

Problem

• Analyzing neuron activation sharing across languages and tasks
• Classifying multilingual neurons into four distinct categories
• Investigating shared neurons' impact on LLM performance degradation

Innovation

• Classifying neurons into four categories across languages
• Deactivating all-shared neurons significantly decreases performance
• Analyzing neuron activation patterns across tasks and languages