Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models

TL;DR

This paper introduces Expert-Specialized Fine-Tuning (ESFT), a method for efficiently fine-tuning sparse-architecture LLMs with Mixture-of-Experts, achieving comparable or better performance than full fine-tuning by focusing on relevant experts.

Contribution

The paper proposes ESFT, a novel PEFT method for MoE-based LLMs that improves efficiency and performance by tuning only task-relevant experts while analyzing expert activation patterns.

Findings

ESFT improves tuning efficiency and performance.

Routing distribution varies across tasks, with concentrated expert activation.

Finer-grained experts in MoE models enhance expert selection and task adaptation.

Abstract

Parameter-efficient fine-tuning (PEFT) is crucial for customizing Large Language Models (LLMs) with constrained resources. Although there have been various PEFT methods for dense-architecture LLMs, PEFT for sparse-architecture LLMs is still underexplored. In this work, we study the PEFT method for LLMs with the Mixture-of-Experts (MoE) architecture and the contents of this work are mainly threefold: (1) We investigate the dispersion degree of the activated experts in customized tasks, and found that the routing distribution for a specific task tends to be highly concentrated, while the distribution of activated experts varies significantly across different tasks. (2) We propose Expert-Specialized Fine-Tuning, or ESFT, which tunes the experts most relevant to downstream tasks while freezing the other experts and modules; experimental results demonstrate that our method not only improves the tuning efficiency, but also matches or even surpasses the performance of full-parameter fine-tuning. (3) We further analyze the impact of the MoE architecture on expert-specialized fine-tuning. We find that MoE models with finer-grained experts are more advantageous in selecting the combination of experts that are most relevant to downstream tasks, thereby enhancing both the training efficiency and effectiveness. Our code is available at https://github.com/deepseek-ai/ESFT.