Metraipolis › archive › dc_cb9b35bad7af

Emergence of Biased Consensus in Multi-Agent LLM Debates — arXiv:2608.02827

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.02827
License (as archived)
CC0-1.0
Archived
2026-10-31T12:14:34.032Z
SHA-256 of the archived text
7850ac546122cfe2591b5b1cf0c39548d0d09263314c32767307c3287e52365c
Stable URL of this page
https://metraipolis.com/documento/dc_cb9b35bad7af

How to cite

Emergence of Biased Consensus in Multi-Agent LLM Debates — arXiv:2608.02827 — arXiv (en), https://arxiv.org/abs/2608.02827 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cb9b35bad7af (Metraipolis archive dc_cb9b35bad7af, sha256 7850ac546122cfe2591b5b1cf0c39548d0d09263314c32767307c3287e52365c)
BibTeX
@misc{dccb9b35bad7af,
  title = {Emergence of Biased Consensus in Multi-Agent LLM Debates — arXiv:2608.02827},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.02827},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cb9b35bad7af; sha256 7850ac546122cfe2591b5b1cf0c39548d0d09263314c32767307c3287e52365c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.02827. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multi-agent LLM debates achieve strong performance on decision-making tasks as well as problem-solving benchmarks, yet their safety and fairness risks remain poorly understood. Notably, interaction can amplify the biases of single LLMs, raising concerns for real-world deployment. We identify the emergence of collective (often biased) norms in multi-agent LLM debates and show that noise (e.g., LLM sampling temperature) is a key driver. To explain this, we propose an analytical framework drawing on physics-inspired theoretical models of social dynamics. We predict a phase transition to collective bias when conformity surpasses a critical threshold given the LLMs' initial bias and debate noise. We test the theoretical predictions through controlled experiments and observe a finite-size crossover consistent with an underlying phase transition. We further find that agent heterogeneity suppresses emergence by smoothing (rounding) this transition. Finally, we show that these insights generalize to realistic decision-making tasks, including investment decisions and LLM-as-a-judge evaluation.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.MA).
Categorie arXiv: Multiagent Systems (cs.MA).
Pagina del lavoro: https://arxiv.org/abs/2608.02827 — 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 2026-10-31T12:05:27.211Z dalla pagina https://arxiv.org/catchup/cs.MA/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).