@inproceedings{feng-etal-2025-sample,
title = "Sample-Efficient Human Evaluation of Large Language Models via Maximum Discrepancy Competition",
author = "Feng, Kehua and
Ding, Keyan and
Hongzhi, Tan and
Ma, Kede and
Wang, Zhihua and
Guo, Shuangquan and
Yuzhou, Cheng and
Sun, Ge and
Zheng, Guozhou and
Zhang, Qiang and
Chen, Huajun",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.535/",
doi = "10.18653/v1/2025.acl-long.535",
pages = "10913--10947",
ISBN = "979-8-89176-251-0",
abstract = "The past years have witnessed a proliferation of large language models (LLMs). Yet, reliable evaluation of LLMs is challenging due to the inaccuracy of standard metrics in human perception of text quality and the inefficiency in sampling informative test examples for human evaluation. This paper presents a sample-efficient human evaluation method for LLMs based on the principle of MAximum Discrepancy (MAD) competition. MAD automatically selects a small set of informative input instructions, each of which maximizes the discrepancy of two LLMs' reponses, which are subsequently subject to three-alternative forced choice by human subjects. The pairwise comparison results of multiple LLMs are then aggregated into a global ranking using the Elo rating system. We compare eight representative LLMs in terms of four skills: knowledge understanding, mathematical reasoning, writing, and coding. Experimental results show that the proposed method reliably achieves the ``golden'' ranking of LLMs with a minimum set of input instructions, which in turn reveal their relative strengths and weaknesses, and offers valuable insights for further LLM advancement."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="feng-etal-2025-sample">
<titleInfo>
<title>Sample-Efficient Human Evaluation of Large Language Models via Maximum Discrepancy Competition</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kehua</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Keyan</namePart>
<namePart type="family">Ding</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tan</namePart>
<namePart type="family">Hongzhi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kede</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhihua</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuangquan</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cheng</namePart>
<namePart type="family">Yuzhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ge</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guozhou</namePart>
<namePart type="family">Zheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qiang</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Huajun</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>The past years have witnessed a proliferation of large language models (LLMs). Yet, reliable evaluation of LLMs is challenging due to the inaccuracy of standard metrics in human perception of text quality and the inefficiency in sampling informative test examples for human evaluation. This paper presents a sample-efficient human evaluation method for LLMs based on the principle of MAximum Discrepancy (MAD) competition. MAD automatically selects a small set of informative input instructions, each of which maximizes the discrepancy of two LLMs’ reponses, which are subsequently subject to three-alternative forced choice by human subjects. The pairwise comparison results of multiple LLMs are then aggregated into a global ranking using the Elo rating system. We compare eight representative LLMs in terms of four skills: knowledge understanding, mathematical reasoning, writing, and coding. Experimental results show that the proposed method reliably achieves the “golden” ranking of LLMs with a minimum set of input instructions, which in turn reveal their relative strengths and weaknesses, and offers valuable insights for further LLM advancement.</abstract>
<identifier type="citekey">feng-etal-2025-sample</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.535</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.535/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>10913</start>
<end>10947</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Sample-Efficient Human Evaluation of Large Language Models via Maximum Discrepancy Competition
%A Feng, Kehua
%A Ding, Keyan
%A Hongzhi, Tan
%A Ma, Kede
%A Wang, Zhihua
%A Guo, Shuangquan
%A Yuzhou, Cheng
%A Sun, Ge
%A Zheng, Guozhou
%A Zhang, Qiang
%A Chen, Huajun
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F feng-etal-2025-sample
%X The past years have witnessed a proliferation of large language models (LLMs). Yet, reliable evaluation of LLMs is challenging due to the inaccuracy of standard metrics in human perception of text quality and the inefficiency in sampling informative test examples for human evaluation. This paper presents a sample-efficient human evaluation method for LLMs based on the principle of MAximum Discrepancy (MAD) competition. MAD automatically selects a small set of informative input instructions, each of which maximizes the discrepancy of two LLMs’ reponses, which are subsequently subject to three-alternative forced choice by human subjects. The pairwise comparison results of multiple LLMs are then aggregated into a global ranking using the Elo rating system. We compare eight representative LLMs in terms of four skills: knowledge understanding, mathematical reasoning, writing, and coding. Experimental results show that the proposed method reliably achieves the “golden” ranking of LLMs with a minimum set of input instructions, which in turn reveal their relative strengths and weaknesses, and offers valuable insights for further LLM advancement.
%R 10.18653/v1/2025.acl-long.535
%U https://aclanthology.org/2025.acl-long.535/
%U https://doi.org/10.18653/v1/2025.acl-long.535
%P 10913-10947
Markdown (Informal)
[Sample-Efficient Human Evaluation of Large Language Models via Maximum Discrepancy Competition](https://aclanthology.org/2025.acl-long.535/) (Feng et al., ACL 2025)
ACL
- Kehua Feng, Keyan Ding, Tan Hongzhi, Kede Ma, Zhihua Wang, Shuangquan Guo, Cheng Yuzhou, Ge Sun, Guozhou Zheng, Qiang Zhang, and Huajun Chen. 2025. Sample-Efficient Human Evaluation of Large Language Models via Maximum Discrepancy Competition. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10913–10947, Vienna, Austria. Association for Computational Linguistics.