Fine-tuning Large Language Models for Improving Factuality in Legal Question Answering

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hu, Yinghao, Gan, Leilei, Xiao, Wenyi, Kuang, Kun, Wu, Fei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916562746736640
author Hu, Yinghao
Gan, Leilei
Xiao, Wenyi
Kuang, Kun
Wu, Fei
author_facet Hu, Yinghao
Gan, Leilei
Xiao, Wenyi
Kuang, Kun
Wu, Fei
contents Hallucination, or the generation of incorrect or fabricated information, remains a critical challenge in large language models (LLMs), particularly in high-stake domains such as legal question answering (QA). In order to mitigate the hallucination rate in legal QA, we first introduce a benchmark called LegalHalBench and three automatic metrics to evaluate the common hallucinations when LLMs answer legal questions. We then propose a hallucination mitigation method that integrates behavior cloning and a novel Hard Sample-aware Iterative Direct Preference Optimization (HIPO). We conduct extensive real-data experiments to validate the effectiveness of our approach. Our results demonstrate remarkable improvements in various metrics, including the newly proposed Non-Hallucinated Statute Rate, Statute Relevance Rate, Legal Claim Truthfulness, as well as traditional metrics such as METEOR, BERTScore, ROUGE-L, and win rates.
format Preprint
id arxiv_https___arxiv_org_abs_2501_06521
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Fine-tuning Large Language Models for Improving Factuality in Legal Question Answering
Hu, Yinghao
Gan, Leilei
Xiao, Wenyi
Kuang, Kun
Wu, Fei
Computation and Language
Hallucination, or the generation of incorrect or fabricated information, remains a critical challenge in large language models (LLMs), particularly in high-stake domains such as legal question answering (QA). In order to mitigate the hallucination rate in legal QA, we first introduce a benchmark called LegalHalBench and three automatic metrics to evaluate the common hallucinations when LLMs answer legal questions. We then propose a hallucination mitigation method that integrates behavior cloning and a novel Hard Sample-aware Iterative Direct Preference Optimization (HIPO). We conduct extensive real-data experiments to validate the effectiveness of our approach. Our results demonstrate remarkable improvements in various metrics, including the newly proposed Non-Hallucinated Statute Rate, Statute Relevance Rate, Legal Claim Truthfulness, as well as traditional metrics such as METEOR, BERTScore, ROUGE-L, and win rates.
title Fine-tuning Large Language Models for Improving Factuality in Legal Question Answering
topic Computation and Language
url https://arxiv.org/abs/2501.06521