RobustFT: Robust Supervised Fine-tuning for Large Language Models under Noisy Response

TL;DR

RobustFT is a novel framework that improves supervised fine-tuning of large language models by effectively detecting and relabeling noisy data, leading to enhanced performance in noisy environments.

Contribution

The paper introduces RobustFT, a noise-robust fine-tuning framework that combines multi-expert noise detection, context-aware relabeling, and entropy-based data selection for large language models.

Findings

RobustFT significantly outperforms baseline methods in noisy data scenarios.

The multi-expert noise detection system improves noise identification accuracy.

Entropy-based sample selection enhances the quality of fine-tuning data.

Abstract

Supervised fine-tuning (SFT) plays a crucial role in adapting large language models (LLMs) to specific domains or tasks. However, as demonstrated by empirical experiments, the collected data inevitably contains noise in practical applications, which poses significant challenges to model performance on downstream tasks. Therefore, there is an urgent need for a noise-robust SFT framework to enhance model capabilities in downstream tasks. To address this challenge, we introduce a robust SFT framework (RobustFT) that performs noise detection and relabeling on downstream task data. For noise identification, our approach employs a multi-expert collaborative system with inference-enhanced models to achieve superior noise detection. In the denoising phase, we utilize a context-enhanced strategy, which incorporates the most relevant and confident knowledge followed by careful assessment to generate reliable annotations. Additionally, we introduce an effective data selection mechanism based on response entropy, ensuring only high-quality samples are retained for fine-tuning. Extensive experiments conducted on multiple LLMs across five datasets demonstrate RobustFT's exceptional performance in noisy scenarios.