Injecting New Knowledge into Large Language Models via Supervised Fine-Tuning

TL;DR

This paper explores supervised fine-tuning methods to effectively inject new, out-of-domain knowledge into large language models, focusing on recent sporting events, and compares dataset generation strategies for improved factual accuracy.

Contribution

It introduces a novel dataset generation process and systematically compares token-based and fact-based scaling for knowledge injection in LLMs.

Findings

Fact-based scaling ensures more uniform knowledge coverage.

Token-based scaling improves Q&A accuracy but lacks coverage consistency.

The proposed method enhances LLM factuality in specific domains.

Abstract

In recent years, Large Language Models (LLMs) have shown remarkable performance in generating human-like text, proving to be a valuable asset across various applications. However, adapting these models to incorporate new, out-of-domain knowledge remains a challenge, particularly for facts and events that occur after the model's knowledge cutoff date. This paper investigates the effectiveness of Supervised Fine-Tuning (SFT) as a method for knowledge injection in LLMs, specifically focusing on the domain of recent sporting events. We compare different dataset generation strategies -- token-based and fact-based scaling -- to create training data that helps the model learn new information. Our experiments on GPT-4 demonstrate that while token-based scaling can lead to improvements in Q&A accuracy, it may not provide uniform coverage of new knowledge. Fact-based scaling, on the other hand, offers a more systematic approach to ensure even coverage across all facts. We present a novel dataset generation process that leads to more effective knowledge ingestion through SFT, and our results show considerable performance improvements in Q&A tasks related to out-of-domain knowledge. This study contributes to the understanding of domain adaptation for LLMs and highlights the potential of SFT in enhancing the factuality of LLM responses in specific knowledge domains.