KaFT: Knowledge-aware Fine-tuning for Boosting LLMs' Domain-specific Question-Answering Performance

TL;DR

This paper introduces KaFT, a novel fine-tuning method that dynamically adjusts training sample weights based on knowledge conflict levels to enhance large language models' domain-specific question-answering performance.

Contribution

The paper proposes a conflict-aware fine-tuning approach that improves LLMs by selectively weighting training data according to knowledge conflict, addressing limitations of traditional supervised fine-tuning.

Findings

KaFT significantly improves QA performance across four LLMs.

Training data with high conflicts can harm model performance if not properly managed.

Appropriate use of conflict data enhances model generalization and reduces hallucinations.

Abstract

Supervised fine-tuning (SFT) is a common approach to improve the domain-specific question-answering (QA) performance of large language models (LLMs). However, recent literature reveals that due to the conflicts between LLMs' internal knowledge and the context knowledge of training data, vanilla SFT using the full QA training set is usually suboptimal. In this paper, we first design a query diversification strategy for robust conflict detection and then conduct a series of experiments to analyze the impact of knowledge conflict. We find that 1) training samples with varied conflicts contribute differently, where SFT on the data with large conflicts leads to catastrophic performance drops; 2) compared to directly filtering out the conflict data, appropriately applying the conflict data would be more beneficial. Motivated by this, we propose a simple-yet-effective Knowledge-aware Fine-tuning (namely KaFT) approach to effectively boost LLMs' performance. The core of KaFT is to adapt the training weight by assigning different rewards for different training samples according to conflict level. Extensive experiments show that KaFT brings consistent and significant improvements across four LLMs. More analyses prove that KaFT effectively improves the model generalization and alleviates the hallucination.