Saved in:
Bibliographic Details
Main Authors: Seo, Jean, Lim, Jongwon, Jang, Dongjun, Shin, Hyopil
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.09255
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915019799658496
author Seo, Jean
Lim, Jongwon
Jang, Dongjun
Shin, Hyopil
author_facet Seo, Jean
Lim, Jongwon
Jang, Dongjun
Shin, Hyopil
contents We introduce DAHL, a benchmark dataset and automated evaluation system designed to assess hallucination in long-form text generation, specifically within the biomedical domain. Our benchmark dataset, meticulously curated from biomedical research papers, consists of 8,573 questions across 29 categories. DAHL evaluates fact-conflicting hallucinations in Large Language Models (LLMs) by deconstructing responses into atomic units, each representing a single piece of information. The accuracy of these responses is averaged to produce the DAHL Score, offering a more in-depth evaluation of hallucinations compared to previous methods that rely on multiple-choice tasks. We conduct experiments with 8 different models, finding that larger models tend to hallucinate less; however, beyond a model size of 7 to 8 billion parameters, further scaling does not significantly improve factual accuracy. The DAHL Score holds potential as an efficient alternative to human-annotated preference labels, being able to be expanded to other specialized domains. We release the dataset and code in public.
format Preprint
id arxiv_https___arxiv_org_abs_2411_09255
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle DAHL: Domain-specific Automated Hallucination Evaluation of Long-Form Text through a Benchmark Dataset in Biomedicine
Seo, Jean
Lim, Jongwon
Jang, Dongjun
Shin, Hyopil
Computation and Language
We introduce DAHL, a benchmark dataset and automated evaluation system designed to assess hallucination in long-form text generation, specifically within the biomedical domain. Our benchmark dataset, meticulously curated from biomedical research papers, consists of 8,573 questions across 29 categories. DAHL evaluates fact-conflicting hallucinations in Large Language Models (LLMs) by deconstructing responses into atomic units, each representing a single piece of information. The accuracy of these responses is averaged to produce the DAHL Score, offering a more in-depth evaluation of hallucinations compared to previous methods that rely on multiple-choice tasks. We conduct experiments with 8 different models, finding that larger models tend to hallucinate less; however, beyond a model size of 7 to 8 billion parameters, further scaling does not significantly improve factual accuracy. The DAHL Score holds potential as an efficient alternative to human-annotated preference labels, being able to be expanded to other specialized domains. We release the dataset and code in public.
title DAHL: Domain-specific Automated Hallucination Evaluation of Long-Form Text through a Benchmark Dataset in Biomedicine
topic Computation and Language
url https://arxiv.org/abs/2411.09255