Selective Self-Rehearsal: A Fine-Tuning Approach to Improve Generalization in Large Language Models

TL;DR

This paper proposes Selective Self-Rehearsal (SSR), a fine-tuning method for large language models that maintains high task performance while significantly improving their ability to generalize to new data, reducing overfitting.

Contribution

SSR introduces a novel fine-tuning approach that uses the model's own correct responses to enhance generalization, outperforming standard supervised fine-tuning in maintaining performance.

Findings

SSR reduces performance drop to around 2% on benchmarks.

SSR outperforms standard fine-tuning in generalization.

Experiments on unanswerable query detection demonstrate effectiveness.

Abstract

Fine-tuning Large Language Models (LLMs) on specific datasets is a common practice to improve performance on target tasks. However, this performance gain often leads to overfitting, where the model becomes too specialized in either the task or the characteristics of the training data, resulting in a loss of generalization. This paper introduces Selective Self-Rehearsal (SSR), a fine-tuning approach that achieves performance comparable to the standard supervised fine-tuning (SFT) while improving generalization. SSR leverages the fact that there can be multiple valid responses to a query. By utilizing the model's correct responses, SSR reduces model specialization during the fine-tuning stage. SSR first identifies the correct model responses from the training set by deploying an appropriate LLM as a judge. Then, it fine-tunes the model using the correct model responses and the gold response for the remaining samples. The effectiveness of SSR is demonstrated through experiments on the task of identifying unanswerable queries across various datasets. The results show that standard SFT can lead to an average performance drop of up to $16.7\%$ on multiple benchmarks, such as MMLU and TruthfulQA. In contrast, SSR results in close to $2\%$ drop on average, indicating better generalization capabilities compared to standard SFT.