Saved in:
Bibliographic Details
Main Authors: Gao, Jun, Peng, Yun, Qiao, Qian, Zhou, Changhai, Zhou, Yuhua, Zhang, Shiyang, Weng, Shichao, Xing, Zhenchang, Ren, Xiaoxue
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.25399
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911628166955008
author Gao, Jun
Peng, Yun
Qiao, Qian
Zhou, Changhai
Zhou, Yuhua
Zhang, Shiyang
Weng, Shichao
Xing, Zhenchang
Ren, Xiaoxue
author_facet Gao, Jun
Peng, Yun
Qiao, Qian
Zhou, Changhai
Zhou, Yuhua
Zhang, Shiyang
Weng, Shichao
Xing, Zhenchang
Ren, Xiaoxue
contents Despite strong performance on code generation tasks, it remains unclear whether large language models (LLMs) genuinely reason about code execution. Existing code reasoning benchmarks primarily evaluate final output correctness under a single canonical implementation, leaving two critical aspects underexplored: (1) whether LLMs can maintain consistency to functionally equivalent implementations, and (2) whether LLMs can accurately reason about intermediate execution states. We introduce \textbf{CoRE}, a \textbf{Co}de \textbf{Re}asoning benchmark that evaluates code reasoning through \textbf{implementation invariance} and \textbf{process transparency}. Extensive evaluations on eight frontier LLMs reveal two fundamental limitations. First, models exhibit a substantial \textbf{robustness gap}, with performance varying significantly across equivalent implementations. Second, we observe \textbf{superficial execution}, where models arrive at correct final outputs without correctly reasoning about intermediate execution states. Together, these findings demonstrate that output-only evaluations are insufficient for assessing code reasoning and position CoRE as a necessary benchmark for evaluating robust and faithful code reasoning.\footnote{Data and code are available at https://github.com/ZJUSig/CoRE.}
format Preprint
id arxiv_https___arxiv_org_abs_2604_25399
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CoRE: A Fine-Grained Code Reasoning Benchmark Beyond Output Prediction
Gao, Jun
Peng, Yun
Qiao, Qian
Zhou, Changhai
Zhou, Yuhua
Zhang, Shiyang
Weng, Shichao
Xing, Zhenchang
Ren, Xiaoxue
Software Engineering
Despite strong performance on code generation tasks, it remains unclear whether large language models (LLMs) genuinely reason about code execution. Existing code reasoning benchmarks primarily evaluate final output correctness under a single canonical implementation, leaving two critical aspects underexplored: (1) whether LLMs can maintain consistency to functionally equivalent implementations, and (2) whether LLMs can accurately reason about intermediate execution states. We introduce \textbf{CoRE}, a \textbf{Co}de \textbf{Re}asoning benchmark that evaluates code reasoning through \textbf{implementation invariance} and \textbf{process transparency}. Extensive evaluations on eight frontier LLMs reveal two fundamental limitations. First, models exhibit a substantial \textbf{robustness gap}, with performance varying significantly across equivalent implementations. Second, we observe \textbf{superficial execution}, where models arrive at correct final outputs without correctly reasoning about intermediate execution states. Together, these findings demonstrate that output-only evaluations are insufficient for assessing code reasoning and position CoRE as a necessary benchmark for evaluating robust and faithful code reasoning.\footnote{Data and code are available at https://github.com/ZJUSig/CoRE.}
title CoRE: A Fine-Grained Code Reasoning Benchmark Beyond Output Prediction
topic Software Engineering
url https://arxiv.org/abs/2604.25399