Metraipolis › archive › dc_653ddd983e37
SpecGuard: Proving a Task Is Broken Before the Agent Cheats — arXiv:2610.09159
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.09159
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T11:09:02.370Z
- SHA-256 of the archived text
6c853de8cb8a6ba4242f54542b42669939c260f13a797bc4bfb23a3111a31f5f- Stable URL of this page
https://metraipolis.com/documento/dc_653ddd983e37
How to cite
SpecGuard: Proving a Task Is Broken Before the Agent Cheats — arXiv:2610.09159 — arXiv (en), https://arxiv.org/abs/2610.09159 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_653ddd983e37 (Metraipolis archive dc_653ddd983e37, sha256 6c853de8cb8a6ba4242f54542b42669939c260f13a797bc4bfb23a3111a31f5f)
BibTeX
@misc{dc653ddd983e37,
title = {SpecGuard: Proving a Task Is Broken Before the Agent Cheats — arXiv:2610.09159},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.09159},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_653ddd983e37; sha256 6c853de8cb8a6ba4242f54542b42669939c260f13a797bc4bfb23a3111a31f5f},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09159. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
As autonomous coding agents get increasingly deployed, the risk that accidental or adversarially injected misspecifications in tasks lead to dangerous agent behavior is critical to address. Prior work has shown that agents given such tasks rarely flag the conflict and instead cheat, editing tests or hard-coding expected outputs, and the actions taken to cheat can cause real damage, such as deleting a security defense to make a corrupted test pass. It remains unclear whether such conflicts can be established with independently verifiable evidence before the agent acts. We present SpecGuard, which detects and formally certifies these conflicts between task intent and tests. Given only the task description and codebase, SpecGuard autoformalizes the intended behaviour into a Lean 4 specification. The tests are formalized independently, and the Lean kernel checks whether any implementation could satisfy both formalizations, producing a machine-checked certificate when none can. On conflicted SWE-bench tasks, SpecGuard detects up to 72.8% of conflicts and formally certifies up to 51.1%, with a nearly five-fold lower conflict miss rate than model-based judgment. SpecGuard provides a pre-execution safety check that identifies reward-hacking opportunities through formal certification of task-level conflicts, before any agent behavior is observed. Our code is available at this https URL . Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI) ; Logic in Computer Science (cs.LO); Software Engineering (cs.SE). Pagina del lavoro: https://arxiv.org/abs/2610.09159 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T10:14:05.911Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).