Metraipolis › archive › dc_19f50264d90d
MADBench: Benchmarking the Security of Multi-Agent Debate — arXiv:2609.39146
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.39146
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T12:42:32.166Z
- SHA-256 of the archived text
fe8b009a1db7f8bae3ce32e4d545fa540b6036d2a0886046701b2f8cc819f8df- Stable URL of this page
https://metraipolis.com/documento/dc_19f50264d90d
How to cite
MADBench: Benchmarking the Security of Multi-Agent Debate — arXiv:2609.39146 — arXiv (en), https://arxiv.org/abs/2609.39146 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_19f50264d90d (Metraipolis archive dc_19f50264d90d, sha256 fe8b009a1db7f8bae3ce32e4d545fa540b6036d2a0886046701b2f8cc819f8df)
BibTeX
@misc{dc19f50264d90d,
title = {MADBench: Benchmarking the Security of Multi-Agent Debate — arXiv:2609.39146},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.39146},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_19f50264d90d; sha256 fe8b009a1db7f8bae3ce32e4d545fa540b6036d2a0886046701b2f8cc819f8df},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39146. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multi-agent debate (MAD) can improve large language model (LLM) reasoning by allowing multiple agents to exchange and critique their answers to the same task. However, the interactions that enable agents to correct mistakes can also spread adversarial errors and steer the agents toward an incorrect answer. Although some efforts have been made to examine particular attack types on MAD, systematic evaluation of MAD under diverse attacks remains limited. A central question is whether debate mitigates adversarial influence or amplifies it. In this paper, we present MADBench, a benchmark for evaluating the security of MAD. We organize attacks into a layered taxonomy following the MAD workflow, incorporating both established attacks and new strategies tailored to debate. We evaluate six attack families over 356 source tasks and 3,958 test cases, examining their effects on the final answer and the propagation of adversarial influence. Our results show that, under attacks, MAD does not necessarily improve LLM reasoning. Compared with a single-agent baseline, MAD can mitigate attacks on answer accuracy in question-answering tasks while amplifying unauthorized reads or writes in both question-answering and workspace tasks. Moreover, even when three out of five agents collude, the attack changes the final answer from correct to wrong on only 28.30\% of tasks answered correctly without attack, while only 3.26\% of initially correct honest agents switch to wrong answers during debate. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.39146 — 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 2026-10-07T12:14:13.737Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).