SurveyEval: Towards Comprehensive Evaluation of LLM-Generated Academic Surveys

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Jiahao, Zhang, Shuaixing, Xu, Nan, Wang, Lei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912743364231168
author Zhao, Jiahao
Zhang, Shuaixing
Xu, Nan
Wang, Lei
author_facet Zhao, Jiahao
Zhang, Shuaixing
Xu, Nan
Wang, Lei
contents LLM-based automatic survey systems are transforming how users acquire information from the web by integrating retrieval, organization, and content synthesis into end-to-end generation pipelines. While recent works focus on developing new generation pipelines, how to evaluate such complex systems remains a significant challenge. To this end, we introduce SurveyEval, a comprehensive benchmark that evaluates automatically generated surveys across three dimensions: overall quality, outline coherence, and reference accuracy. We extend the evaluation across 7 subjects and augment the LLM-as-a-Judge framework with human references to strengthen evaluation-human alignment. Evaluation results show that while general long-text or paper-writing systems tend to produce lower-quality surveys, specialized survey-generation systems are able to deliver substantially higher-quality results. We envision SurveyEval as a scalable testbed to understand and improve automatic survey systems across diverse subjects and evaluation criteria.
format Preprint
id arxiv_https___arxiv_org_abs_2512_02763
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SurveyEval: Towards Comprehensive Evaluation of LLM-Generated Academic Surveys
Zhao, Jiahao
Zhang, Shuaixing
Xu, Nan
Wang, Lei
Computation and Language
Artificial Intelligence
LLM-based automatic survey systems are transforming how users acquire information from the web by integrating retrieval, organization, and content synthesis into end-to-end generation pipelines. While recent works focus on developing new generation pipelines, how to evaluate such complex systems remains a significant challenge. To this end, we introduce SurveyEval, a comprehensive benchmark that evaluates automatically generated surveys across three dimensions: overall quality, outline coherence, and reference accuracy. We extend the evaluation across 7 subjects and augment the LLM-as-a-Judge framework with human references to strengthen evaluation-human alignment. Evaluation results show that while general long-text or paper-writing systems tend to produce lower-quality surveys, specialized survey-generation systems are able to deliver substantially higher-quality results. We envision SurveyEval as a scalable testbed to understand and improve automatic survey systems across diverse subjects and evaluation criteria.
title SurveyEval: Towards Comprehensive Evaluation of LLM-Generated Academic Surveys
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2512.02763