@inproceedings{yuan-etal-2024-poor,
title = "Poor-Supervised Evaluation for {S}uper{LLM} via Mutual Consistency",
author = "Yuan, Peiwen and
Feng, Shaoxiong and
Li, Yiwei and
Wang, Xinglin and
Pan, Boyuan and
Wang, Heda and
Hu, Yao and
Li, Kan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.690",
doi = "10.18653/v1/2024.findings-acl.690",
pages = "11614--11627",
abstract = "The guidance from capability evaluations has greatly propelled the progress of human society and the development of Artificial Intelligence. However, as LLMs evolve, it becomes challenging to construct evaluation benchmark with accurate labels for SuperLLMs whose capabilities approach or even surpass those of humans. To credibly conduct poor-supervised evaluation without accurate labels, we first prove that the consistency between the model under evaluation and the reference model, when their prediction distributions are independent and the sample size is infinite, can equivalently assess the true capabilities of the model to be evaluated. However, using either humans or LLMs as the reference model cannot sufficiently meet the conditions, for which we propose the PEEM algorithm. By treating all models under evaluation as reference models, PEEM alternately optimizes model weights and filters reference models based on EM algorithm to maximally alleviate the insufficiency of the conditions. Comprehensive experiments across 3 types of tasks with 16 mainstream LLMs validate the efficiency, universality, and effectiveness of PEEM. More generally, PEEM has advanced the evaluation paradigm evolution from human-centric to human{\&}model-centric, alleviating the limitations of human capabilities for evaluating SuperLLMs.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yuan-etal-2024-poor">
<titleInfo>
<title>Poor-Supervised Evaluation for SuperLLM via Mutual Consistency</title>
</titleInfo>
<name type="personal">
<namePart type="given">Peiwen</namePart>
<namePart type="family">Yuan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shaoxiong</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yiwei</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinglin</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Boyuan</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Heda</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yao</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lun-Wei</namePart>
<namePart type="family">Ku</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Martins</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Srikumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The guidance from capability evaluations has greatly propelled the progress of human society and the development of Artificial Intelligence. However, as LLMs evolve, it becomes challenging to construct evaluation benchmark with accurate labels for SuperLLMs whose capabilities approach or even surpass those of humans. To credibly conduct poor-supervised evaluation without accurate labels, we first prove that the consistency between the model under evaluation and the reference model, when their prediction distributions are independent and the sample size is infinite, can equivalently assess the true capabilities of the model to be evaluated. However, using either humans or LLMs as the reference model cannot sufficiently meet the conditions, for which we propose the PEEM algorithm. By treating all models under evaluation as reference models, PEEM alternately optimizes model weights and filters reference models based on EM algorithm to maximally alleviate the insufficiency of the conditions. Comprehensive experiments across 3 types of tasks with 16 mainstream LLMs validate the efficiency, universality, and effectiveness of PEEM. More generally, PEEM has advanced the evaluation paradigm evolution from human-centric to human&model-centric, alleviating the limitations of human capabilities for evaluating SuperLLMs.</abstract>
<identifier type="citekey">yuan-etal-2024-poor</identifier>
<identifier type="doi">10.18653/v1/2024.findings-acl.690</identifier>
<location>
<url>https://aclanthology.org/2024.findings-acl.690</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>11614</start>
<end>11627</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Poor-Supervised Evaluation for SuperLLM via Mutual Consistency
%A Yuan, Peiwen
%A Feng, Shaoxiong
%A Li, Yiwei
%A Wang, Xinglin
%A Pan, Boyuan
%A Wang, Heda
%A Hu, Yao
%A Li, Kan
%Y Ku, Lun-Wei
%Y Martins, Andre
%Y Srikumar, Vivek
%S Findings of the Association for Computational Linguistics: ACL 2024
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F yuan-etal-2024-poor
%X The guidance from capability evaluations has greatly propelled the progress of human society and the development of Artificial Intelligence. However, as LLMs evolve, it becomes challenging to construct evaluation benchmark with accurate labels for SuperLLMs whose capabilities approach or even surpass those of humans. To credibly conduct poor-supervised evaluation without accurate labels, we first prove that the consistency between the model under evaluation and the reference model, when their prediction distributions are independent and the sample size is infinite, can equivalently assess the true capabilities of the model to be evaluated. However, using either humans or LLMs as the reference model cannot sufficiently meet the conditions, for which we propose the PEEM algorithm. By treating all models under evaluation as reference models, PEEM alternately optimizes model weights and filters reference models based on EM algorithm to maximally alleviate the insufficiency of the conditions. Comprehensive experiments across 3 types of tasks with 16 mainstream LLMs validate the efficiency, universality, and effectiveness of PEEM. More generally, PEEM has advanced the evaluation paradigm evolution from human-centric to human&model-centric, alleviating the limitations of human capabilities for evaluating SuperLLMs.
%R 10.18653/v1/2024.findings-acl.690
%U https://aclanthology.org/2024.findings-acl.690
%U https://doi.org/10.18653/v1/2024.findings-acl.690
%P 11614-11627
Markdown (Informal)
[Poor-Supervised Evaluation for SuperLLM via Mutual Consistency](https://aclanthology.org/2024.findings-acl.690) (Yuan et al., Findings 2024)
ACL
- Peiwen Yuan, Shaoxiong Feng, Yiwei Li, Xinglin Wang, Boyuan Pan, Heda Wang, Yao Hu, and Kan Li. 2024. Poor-Supervised Evaluation for SuperLLM via Mutual Consistency. In Findings of the Association for Computational Linguistics: ACL 2024, pages 11614–11627, Bangkok, Thailand. Association for Computational Linguistics.