Saved in:
Bibliographic Details
Main Authors: Heineking, Sebastian, Probst, Jonas, Steinbach, Daniel, Potthast, Martin, Scells, Harrisen
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.09831
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912191420039168
author Heineking, Sebastian
Probst, Jonas
Steinbach, Daniel
Potthast, Martin
Scells, Harrisen
author_facet Heineking, Sebastian
Probst, Jonas
Steinbach, Daniel
Potthast, Martin
Scells, Harrisen
contents Evaluating the output of generative large language models (LLMs) is challenging and difficult to scale. Many evaluations of LLMs focus on tasks such as single-choice question-answering or text classification. These tasks are not suitable for assessing open-ended question-answering capabilities, which are critical in domains where expertise is required. One such domain is health, where misleading or incorrect answers can have a negative impact on a user's well-being. Using human experts to evaluate the quality of LLM answers is generally considered the gold standard, but expert annotation is costly and slow. We present a method for evaluating LLM answers that uses ranking models trained on annotated document collections as a substitute for explicit relevance judgements and apply it to the CLEF 2021 eHealth dataset. In a user study, our method correlates with the preferences of a human expert (Kendall's $τ=0.64$). It is also consistent with previous findings in that the quality of generated answers improves with the size of the model and more sophisticated prompting strategies.
format Preprint
id arxiv_https___arxiv_org_abs_2408_09831
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Ranking Generated Answers: On the Agreement of Retrieval Models with Humans on Consumer Health Questions
Heineking, Sebastian
Probst, Jonas
Steinbach, Daniel
Potthast, Martin
Scells, Harrisen
Information Retrieval
Evaluating the output of generative large language models (LLMs) is challenging and difficult to scale. Many evaluations of LLMs focus on tasks such as single-choice question-answering or text classification. These tasks are not suitable for assessing open-ended question-answering capabilities, which are critical in domains where expertise is required. One such domain is health, where misleading or incorrect answers can have a negative impact on a user's well-being. Using human experts to evaluate the quality of LLM answers is generally considered the gold standard, but expert annotation is costly and slow. We present a method for evaluating LLM answers that uses ranking models trained on annotated document collections as a substitute for explicit relevance judgements and apply it to the CLEF 2021 eHealth dataset. In a user study, our method correlates with the preferences of a human expert (Kendall's $τ=0.64$). It is also consistent with previous findings in that the quality of generated answers improves with the size of the model and more sophisticated prompting strategies.
title Ranking Generated Answers: On the Agreement of Retrieval Models with Humans on Consumer Health Questions
topic Information Retrieval
url https://arxiv.org/abs/2408.09831