Mitigating Hallucination in Fictional Character Role-Play

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sadeq, Nafis, Xie, Zhouhang, Kang, Byungkyu, Lamba, Prarit, Gao, Xiang, McAuley, Julian
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910691407953920
author Sadeq, Nafis
Xie, Zhouhang
Kang, Byungkyu
Lamba, Prarit
Gao, Xiang
McAuley, Julian
author_facet Sadeq, Nafis
Xie, Zhouhang
Kang, Byungkyu
Lamba, Prarit
Gao, Xiang
McAuley, Julian
contents Role-playing has wide-ranging applications in customer support, embodied agents, and computational social science. The influence of parametric world knowledge of large language models (LLMs) often causes role-playing characters to act out of character and to hallucinate about things outside the scope of their knowledge. In this work, we focus on the evaluation and mitigation of hallucination in fictional character role-play. We introduce a dataset with over 2,000 characters and 72,000 interviews, including 18,000 adversarial questions. We propose RoleFact, a role-playing method that mitigates hallucination by modulating the influence of parametric knowledge using a pre-calibrated confidence threshold. Experiments show that the proposed method improves the factual precision of generated responses by 18% for adversarial questions with a 44% reduction in temporal hallucination for time-sensitive interviews. The code and the dataset are available at https://github.com/NafisSadeq/rolefact.git.
format Preprint
id arxiv_https___arxiv_org_abs_2406_17260
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Mitigating Hallucination in Fictional Character Role-Play
Sadeq, Nafis
Xie, Zhouhang
Kang, Byungkyu
Lamba, Prarit
Gao, Xiang
McAuley, Julian
Computation and Language
Role-playing has wide-ranging applications in customer support, embodied agents, and computational social science. The influence of parametric world knowledge of large language models (LLMs) often causes role-playing characters to act out of character and to hallucinate about things outside the scope of their knowledge. In this work, we focus on the evaluation and mitigation of hallucination in fictional character role-play. We introduce a dataset with over 2,000 characters and 72,000 interviews, including 18,000 adversarial questions. We propose RoleFact, a role-playing method that mitigates hallucination by modulating the influence of parametric knowledge using a pre-calibrated confidence threshold. Experiments show that the proposed method improves the factual precision of generated responses by 18% for adversarial questions with a 44% reduction in temporal hallucination for time-sensitive interviews. The code and the dataset are available at https://github.com/NafisSadeq/rolefact.git.
title Mitigating Hallucination in Fictional Character Role-Play
topic Computation and Language
url https://arxiv.org/abs/2406.17260