Towards Understanding Multi-Task Learning (Generalization) of LLMs via Detecting and Exploring Task-Specific Neurons

TL;DR

This paper investigates how task-specific neurons in large language models influence multi-task learning and generalization, revealing neuron overlaps' importance and proposing a neuron-level fine-tuning method to improve continual learning.

Contribution

It introduces a method to detect task-specific neurons in LLMs and demonstrates their role in generalization and catastrophic forgetting, proposing a neuron-level fine-tuning approach.

Findings

Task-specific neurons are highly correlated with specific tasks.

Overlap of task-specific neurons relates to better generalization.

Neuron-level fine-tuning improves continual learning performance.

Abstract

While large language models (LLMs) have demonstrated superior multi-task capabilities, understanding the learning mechanisms behind this is still a challenging problem. In this paper, we attempt to understand such mechanisms from the perspective of neurons. Specifically, we detect task-sensitive neurons in LLMs via gradient attribution on task-specific data. Through extensive deactivation and fine-tuning experiments, we demonstrate that the detected neurons are highly correlated with the given task, which we term as task-specific neurons. With these identified task-specific neurons, we delve into two common problems in multi-task learning and continuous learning: Generalization and Catastrophic Forgetting. We find that the overlap of task-specific neurons is strongly associated with generalization and specialization across tasks. Interestingly, at certain layers of LLMs, there is a high similarity in the parameters of different task-specific neurons, and such similarity is highly correlated with the generalization performance. Inspired by these findings, we propose a neuron-level continuous fine-tuning method that only fine-tunes the current task-specific neurons during continuous learning, and extensive experiments demonstrate the effectiveness of the proposed method. Our study provides insights into the interpretability of LLMs in multi-task learning.