Automatically Benchmarking LLM Code Agents through Agent-Driven Annotation and Evaluation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fu, Lingyue, Zhang, Bolun, Guan, Hao, Zhu, Yaoming, Qiu, Lin, Liu, Weiwen, Cao, Xuezhi, Cai, Xunliang, Zhang, Weinan, Yu, Yong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914414063517696
author Fu, Lingyue
Zhang, Bolun
Guan, Hao
Zhu, Yaoming
Qiu, Lin
Liu, Weiwen
Cao, Xuezhi
Cai, Xunliang
Zhang, Weinan
Yu, Yong
author_facet Fu, Lingyue
Zhang, Bolun
Guan, Hao
Zhu, Yaoming
Qiu, Lin
Liu, Weiwen
Cao, Xuezhi
Cai, Xunliang
Zhang, Weinan
Yu, Yong
contents Recent advances in code agents have enabled automated software development at the project level, supported by large language models (LLMs). However, existing benchmarks for code agent evaluation face two major limitations. First, creating high-quality project-level evaluation datasets requires extensive domain expertise, leading to prohibitive annotation costs and limited diversity. Second, while recent Agent-as-a-Judge paradigms address the rigidity of traditional unit tests by enabling flexible metrics, their reliance on In-Context Learning (ICL) with general LLMs often results in inaccurate assessments that misalign with human standards. To address these challenges, we propose an agent-driven benchmark construction pipeline that leverages human supervision to efficiently generate diverse project-level tasks. Based on this, we introduce PRDBench, comprising 50 real-world Python projects across 20 domains, each with structured Product Requirement Documents (PRDs) and comprehensive criteria. Furthermore, to overcome the inaccuracy of general LLM judges, we propose a highly reliable evaluation framework powered by a specialized, fine-tuned model. Based on Qwen3-Coder-30B, our dedicated PRDJudge achieves over 90% human alignment in fixed-interface scenarios. Extensive experiments demonstrate that our suite provides a scalable, robust, and highly accurate framework for assessing state-of-the-art code agents.
format Preprint
id arxiv_https___arxiv_org_abs_2510_24358
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Automatically Benchmarking LLM Code Agents through Agent-Driven Annotation and Evaluation
Fu, Lingyue
Zhang, Bolun
Guan, Hao
Zhu, Yaoming
Qiu, Lin
Liu, Weiwen
Cao, Xuezhi
Cai, Xunliang
Zhang, Weinan
Yu, Yong
Software Engineering
Computation and Language
Recent advances in code agents have enabled automated software development at the project level, supported by large language models (LLMs). However, existing benchmarks for code agent evaluation face two major limitations. First, creating high-quality project-level evaluation datasets requires extensive domain expertise, leading to prohibitive annotation costs and limited diversity. Second, while recent Agent-as-a-Judge paradigms address the rigidity of traditional unit tests by enabling flexible metrics, their reliance on In-Context Learning (ICL) with general LLMs often results in inaccurate assessments that misalign with human standards. To address these challenges, we propose an agent-driven benchmark construction pipeline that leverages human supervision to efficiently generate diverse project-level tasks. Based on this, we introduce PRDBench, comprising 50 real-world Python projects across 20 domains, each with structured Product Requirement Documents (PRDs) and comprehensive criteria. Furthermore, to overcome the inaccuracy of general LLM judges, we propose a highly reliable evaluation framework powered by a specialized, fine-tuned model. Based on Qwen3-Coder-30B, our dedicated PRDJudge achieves over 90% human alignment in fixed-interface scenarios. Extensive experiments demonstrate that our suite provides a scalable, robust, and highly accurate framework for assessing state-of-the-art code agents.
title Automatically Benchmarking LLM Code Agents through Agent-Driven Annotation and Evaluation
topic Software Engineering
Computation and Language
url https://arxiv.org/abs/2510.24358