Exploring Knowledge Purification in Multi-Teacher Knowledge Distillation for LLMs

TL;DR

This paper introduces Knowledge Purification, a technique to consolidate multiple teacher LLMs' rationales into one, reducing conflicts and improving efficiency in knowledge distillation.

Contribution

It proposes five purification methods to enhance multi-teacher knowledge distillation and demonstrates their effectiveness in improving model performance.

Findings

Purification methods improve distilled model performance.

They effectively reduce knowledge conflicts.

Router-based methods show strong generalization.

Abstract

Knowledge distillation has emerged as a pivotal technique for transferring knowledge from stronger large language models (LLMs) to smaller, more efficient models. However, traditional distillation approaches face challenges related to knowledge conflicts and high resource demands, particularly when leveraging multiple teacher models. In this paper, we introduce the concept of \textbf{Knowledge Purification}, which consolidates the rationales from multiple teacher LLMs into a single rationale, thereby mitigating conflicts and enhancing efficiency. To investigate the effectiveness of knowledge purification, we further propose five purification methods from various perspectives. Our experiments demonstrate that these methods not only improve the performance of the distilled model but also effectively alleviate knowledge conflicts. Moreover, router-based methods exhibit robust generalization capabilities, underscoring the potential of innovative purification techniques in optimizing multi-teacher distillation and facilitating the practical deployment of powerful yet lightweight models.