ClassEval-Pro: A Cross-Domain Benchmark for Class-Level Code Generation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Yeheng, Xie, Chaoxiang, Shi, Yuling, Zeng, Wenhao, Wang, Yongpan, Zhang, Hongyu, Gu, Xiaodong
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909000975515648
author Chen, Yeheng
Xie, Chaoxiang
Shi, Yuling
Zeng, Wenhao
Wang, Yongpan
Zhang, Hongyu
Gu, Xiaodong
author_facet Chen, Yeheng
Xie, Chaoxiang
Shi, Yuling
Zeng, Wenhao
Wang, Yongpan
Zhang, Hongyu
Gu, Xiaodong
contents LLMs have achieved strong results on both function-level code synthesis and repository-level code modification, yet a capability that falls between these two extremes -- compositional code creation, i.e., building a complete, internally structured class from a specification -- remains underserved. Current evaluations are either confined to isolated functions or rely on manually curated class-level tasks that are expensive to scale and increasingly susceptible to data contamination. We introduce ClassEval-Pro, a benchmark of 300 class-level tasks spanning 11 domains, constructed through an automated three-stage pipeline that combines complexity enhancement, cross-domain class composition, and integration of real-world GitHub code contributed after January 2025. Every task is validated by an LLM Judge Ensemble and must pass test suites with over 90% line coverage. We evaluate five frontier LLMs under five generation strategies. The best model achieves only 45.6% class-level Pass@1, with a 17.7-point gap between the strongest and weakest models, confirming the benchmark's discriminative power. Strategy choice strongly interacts with model capability: structured approaches such as bottom-up improve weaker models by up to 9.4 percentage points, while compositional generation collapses to as low as 1.3%. Error analysis over 500 manually annotated failures reveals that logic errors (56.2%) and dependency errors (38.0%) dominate, identifying cross-method coordination as the core bottleneck.
format Preprint
id arxiv_https___arxiv_org_abs_2604_26923
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ClassEval-Pro: A Cross-Domain Benchmark for Class-Level Code Generation
Chen, Yeheng
Xie, Chaoxiang
Shi, Yuling
Zeng, Wenhao
Wang, Yongpan
Zhang, Hongyu
Gu, Xiaodong
Software Engineering
Computation and Language
LLMs have achieved strong results on both function-level code synthesis and repository-level code modification, yet a capability that falls between these two extremes -- compositional code creation, i.e., building a complete, internally structured class from a specification -- remains underserved. Current evaluations are either confined to isolated functions or rely on manually curated class-level tasks that are expensive to scale and increasingly susceptible to data contamination. We introduce ClassEval-Pro, a benchmark of 300 class-level tasks spanning 11 domains, constructed through an automated three-stage pipeline that combines complexity enhancement, cross-domain class composition, and integration of real-world GitHub code contributed after January 2025. Every task is validated by an LLM Judge Ensemble and must pass test suites with over 90% line coverage. We evaluate five frontier LLMs under five generation strategies. The best model achieves only 45.6% class-level Pass@1, with a 17.7-point gap between the strongest and weakest models, confirming the benchmark's discriminative power. Strategy choice strongly interacts with model capability: structured approaches such as bottom-up improve weaker models by up to 9.4 percentage points, while compositional generation collapses to as low as 1.3%. Error analysis over 500 manually annotated failures reveals that logic errors (56.2%) and dependency errors (38.0%) dominate, identifying cross-method coordination as the core bottleneck.
title ClassEval-Pro: A Cross-Domain Benchmark for Class-Level Code Generation
topic Software Engineering
Computation and Language
url https://arxiv.org/abs/2604.26923