Metraipolis › archive › dc_997b81fc3383

LEANGUARD: Lean-Verified Guardrails for Tool-Using Agents — arXiv:2611.02332

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.02332
License (as archived)
CC0-1.0
Archived
2026-11-03T18:13:55.806Z
SHA-256 of the archived text
fbbe7bbd832b92ec351f5f6e3fe1167ea212421454b2747cba15a8f3ae0e0213
Stable URL of this page
https://metraipolis.com/documento/dc_997b81fc3383

How to cite

LEANGUARD: Lean-Verified Guardrails for Tool-Using Agents — arXiv:2611.02332 — arXiv (en), https://arxiv.org/abs/2611.02332 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_997b81fc3383 (Metraipolis archive dc_997b81fc3383, sha256 fbbe7bbd832b92ec351f5f6e3fe1167ea212421454b2747cba15a8f3ae0e0213)
BibTeX
@misc{dc997b81fc3383,
  title = {LEANGUARD: Lean-Verified Guardrails for Tool-Using Agents — arXiv:2611.02332},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.02332},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_997b81fc3383; sha256 fbbe7bbd832b92ec351f5f6e3fe1167ea212421454b2747cba15a8f3ae0e0213},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.02332. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Tool-using language-model agents often execute actions that their domain policies forbid. We present LeanGuard, a runtime that checks every proposed tool call against a policy written in Lean4 before the call reaches the environment. Policies combine state predicates with scoped past-time temporal operators over the execution history, evaluation is fail-closed, and the decision procedure is proved equivalent to a LeanLTL semantics. A denied call can be corrected in two ways. With feedback, the agent receives the violated rule and a hint; with proof-guided repair, Lean metaprogramming searches for a nearby call that preserves the user's intent and proves that the policy admits it, without the model. On Tau-Bench with five Qwen models (0.8B-27B), unguarded agents execute at least 33 calls that the Lean policy forbids, while LeanGuard executes none; with feedback, task success does not change significantly. On DTap, LeanGuard reduces direct prompt-injection success from 57.9% to 5.9%, and in telecom, where benign tasks need no writes, benign success is unchanged. Feedback reduces repeated blocked calls and raises task success for four of five models over a bare denial. On replayed denials, repair proves an admissible edit for 88 of 252 denials caused by argument values, in a median of 2,ms.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2611.02332 — 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 2026-11-03T18:09:53.383Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).