Recently, the ICSE 2026 Program Committee announced the recipients of the ACM SIGSOFT Distinguished Paper Awards. The paper titled "HoarePrompt: Structural Reasoning About Program Correctness in Natural Language," published at ICSE '26 by Sergey Mechtaev's team, was honored with this award. Established by ACM SIGSOFT, the award recognizes the top 10% of outstanding papers presented at premier software engineering conferences.

The first author of the paper is Dimitris Bouras, a Class of 2025 PhD student from Peking University (advised by Sergey Mechtaev). The corresponding author is Assistant Professor Sergey Mechtaev from Peking University.

This paper studies whether large language models can determine if a program is correct with respect to a natural-language specification. The authors observe that directly asking an LLM to judge correctness performs poorly because the model struggles to reason about program semantics. HoarePrompt addresses this by structuring the reasoning process: it propagates natural-language descriptions of reachable program states through the code, effectively constructing informal Hoare-style reasoning traces that summarize program behavior. The annotated program is then used by the LLM to decide whether the implementation satisfies the specification. This structured reasoning substantially improves correctness classification and demonstrates that natural-language state reasoning can serve as a practical bridge between informal specifications and program analysis.