Metraipolis › archive › dc_7206ce002019
Who Convinced Whom? Tracing How LLMs Change Their Minds in Multi-Agent Systems — arXiv:2610.35930
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.35930
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-28T03:01:03.594Z
- SHA-256 of the archived text
1817199f61159d2cda8151ff3adf21d9e0a1a168f0d915eb4079a0a27559db4d- Stable URL of this page
https://metraipolis.com/documento/dc_7206ce002019
How to cite
Who Convinced Whom? Tracing How LLMs Change Their Minds in Multi-Agent Systems — arXiv:2610.35930 — arXiv (en), https://arxiv.org/abs/2610.35930 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7206ce002019 (Metraipolis archive dc_7206ce002019, sha256 1817199f61159d2cda8151ff3adf21d9e0a1a168f0d915eb4079a0a27559db4d)
BibTeX
@misc{dc7206ce002019,
title = {Who Convinced Whom? Tracing How LLMs Change Their Minds in Multi-Agent Systems — arXiv:2610.35930},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.35930},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7206ce002019; sha256 1817199f61159d2cda8151ff3adf21d9e0a1a168f0d915eb4079a0a27559db4d},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.35930. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multi-agent debate (MAD) aims to improve language-model decisions through the exchange of arguments, yet agreement after debate need not reflect independent corroboration. We investigate how peer communication influences individual decisions and collective outcomes through behavioral and internal interventions. We introduce MAD Bench, a 2,400-question benchmark spanning knowledge and reasoning, values and moral judgment, and safety and refusal, and evaluate three open-weight models using a round-matched independent-reflection control. Among answer-changing responses, a forced-choice readout before rationale generation matches the eventual answer more often in debate than in reflection (65.4% versus 22.3% on Qwen3.5-9B). Message ablations reveal model-dependent effects of peer answers and reasoning, while attributing identical text to the receiver rather than another agent substantially changes its influence. For the tested source descriptions, stated track record affects answer preferences more strongly than professional-status cues across all three models. Attention masking produces different changes in decision margins across the three checkpoints. On fully parsed, reference-labelled non-numeric questions, debate yields small estimated accuracy improvements over initial majority voting, while increasing the rate of unanimous wrong answers by 3.1-7.4 percentage points relative to reflection. Longer debates show persistent consensus and greater correlation between agents' correctness than reflection. These findings show that stronger consensus can coexist with more shared errors, motivating evaluations that assess dependence and unanimous errors alongside accuracy. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.MA, nuovi lavori del giorno). Categorie arXiv: Multiagent Systems (cs.MA) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.35930 — 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 2026-10-28T03:01:03.513Z dalla pagina https://arxiv.org/list/cs.MA/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).