Research

Paper

AI LLM March 23, 2026

Lemma Discovery in Agentic Program Verification

Authors

Huan Zhao, Haoxin Tu, Zhengyao Liu, Martin Rinard, Abhik Roychoudhury

Abstract

Deductive verification provides strong correctness guarantees for code by extracting verification conditions (VCs) and writing formal proofs for them. The expertise-intensive task of VC proving is the main bottleneck in this process, and has been partly automated owing to recent advances in Large Language Model (LLM) agents. However, existing proof agents are not able to discover helper lemmas - auxiliary lemmas that aid in proving - and thus fall short as programs grow in size and complexity. In this paper, we argue that VC proving for program verification is more than a purely mathematical task, and benefits considerably from program comprehension. Our key insight is that human-proof engineers often discover and apply helper lemmas based on their understanding of the program semantics, which are not directly reflected in the VCs produced by VC generators. Inspired by this insight, we propose an LLM agent, LemmaNet, that discovers helper lemmas in two ways. Specifically, the agent first synthesizes lemmas offline by directly analyzing the source code and specifications, and then relating this semantic understanding to the mechanical, verbose encoding produced by VC generators. As the proof unfolds, LemmaNet then adapts existing helper lemmas online to accommodate evolving proof states, enabling the agent to effectively discharge complex VCs on-the-fly. We evaluate LemmaNet on SV-COMP and established real-world subjects, including modules of the Linux kernel, Contiki OS, standard C++ library, and X.509 parser. Our experimental results demonstrate that LemmaNet significantly outperforms state-of-the-art approaches, highlighting the importance of program comprehension-aided lemma discovery in agentic program verification.

Metadata

arXiv ID: 2603.22114
Provider: ARXIV
Primary Category: cs.SE
Published: 2026-03-23
Fetched: 2026-03-24 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.22114v1</id>\n    <title>Lemma Discovery in Agentic Program Verification</title>\n    <updated>2026-03-23T15:42:07Z</updated>\n    <link href='https://arxiv.org/abs/2603.22114v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.22114v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Deductive verification provides strong correctness guarantees for code by extracting verification conditions (VCs) and writing formal proofs for them. The expertise-intensive task of VC proving is the main bottleneck in this process, and has been partly automated owing to recent advances in Large Language Model (LLM) agents. However, existing proof agents are not able to discover helper lemmas - auxiliary lemmas that aid in proving - and thus fall short as programs grow in size and complexity.\n  In this paper, we argue that VC proving for program verification is more than a purely mathematical task, and benefits considerably from program comprehension. Our key insight is that human-proof engineers often discover and apply helper lemmas based on their understanding of the program semantics, which are not directly reflected in the VCs produced by VC generators. Inspired by this insight, we propose an LLM agent, LemmaNet, that discovers helper lemmas in two ways. Specifically, the agent first synthesizes lemmas offline by directly analyzing the source code and specifications, and then relating this semantic understanding to the mechanical, verbose encoding produced by VC generators. As the proof unfolds, LemmaNet then adapts existing helper lemmas online to accommodate evolving proof states, enabling the agent to effectively discharge complex VCs on-the-fly.\n  We evaluate LemmaNet on SV-COMP and established real-world subjects, including modules of the Linux kernel, Contiki OS, standard C++ library, and X.509 parser. Our experimental results demonstrate that LemmaNet significantly outperforms state-of-the-art approaches, highlighting the importance of program comprehension-aided lemma discovery in agentic program verification.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.SE'/>\n    <published>2026-03-23T15:42:07Z</published>\n    <arxiv:comment>12 pages, 9 figures, 3 tables</arxiv:comment>\n    <arxiv:primary_category term='cs.SE'/>\n    <author>\n      <name>Huan Zhao</name>\n    </author>\n    <author>\n      <name>Haoxin Tu</name>\n    </author>\n    <author>\n      <name>Zhengyao Liu</name>\n    </author>\n    <author>\n      <name>Martin Rinard</name>\n    </author>\n    <author>\n      <name>Abhik Roychoudhury</name>\n    </author>\n  </entry>"
}