Metraipolis › archive › dc_10d6e697e10d
POLAR: Ontology-Guided Risk Prevention for Tool-Calling LLM Agents — arXiv:2610.08082
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.08082
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T11:19:24.307Z
- SHA-256 of the archived text
d3c423fb5c720a7135d790e50e7570b0c361724b5c4ca38b5b508055cdbbf6b2- Stable URL of this page
https://metraipolis.com/documento/dc_10d6e697e10d
How to cite
POLAR: Ontology-Guided Risk Prevention for Tool-Calling LLM Agents — arXiv:2610.08082 — arXiv (en), https://arxiv.org/abs/2610.08082 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_10d6e697e10d (Metraipolis archive dc_10d6e697e10d, sha256 d3c423fb5c720a7135d790e50e7570b0c361724b5c4ca38b5b508055cdbbf6b2)
BibTeX
@misc{dc10d6e697e10d,
title = {POLAR: Ontology-Guided Risk Prevention for Tool-Calling LLM Agents — arXiv:2610.08082},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.08082},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_10d6e697e10d; sha256 d3c423fb5c720a7135d790e50e7570b0c361724b5c4ca38b5b508055cdbbf6b2},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.08082. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM tool-use agents operate in dynamic environments where many actions carry operational risk. However, most safety mechanisms react only after errors manifest. Existing pre-emptive approaches either fine-tune the agent on chain-of-thought deliberation or compile natural-language guardrails into runtime checks, but they do so without exposing a structural, auditable verdict. We propose POLAR, a guardrail framework for small tool-calling agents that assesses reversibility through a structured two-layer ontology. POLAR assigns each action a graded reversibility score by deriving a candidate inverse sequence; calls failing a threshold are pruned before execution. Evaluated on $\tau^2$-bench across six agent models, POLAR improves mean task reward by 0.11 to 0.18 points on airline for four of six agents, but only eight of eighteen model--domain cells improve overall; retail and stronger agents often regress. POLAR provides an auditable structural check and characterizes its task-utility trade-offs. Reward is not a direct measure of prevented harm. 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/2610.08082 — 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 2026-10-07T11:15:24.581Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).