pith. sign in

arxiv: 2605.25929 · v2 · pith:6BJ6FVG7new · submitted 2026-05-25 · 💻 cs.MA · cs.LG

Multi-Agent Systems are Mixtures of Experts: Who Becomes an Influencer?

Franka Bause , Jonas Niederle , Martin Pawelczyk , Rebekka Burkholz This is my paper

Pith reviewed 2026-06-29 19:24 UTC · model grok-4.3

classification 💻 cs.MA cs.LG
keywords multi-agent systemsLLM deliberationopinion dynamicsmixture of expertsinfluence mechanismsFriedkin-Johnsen modelagent competence
0
0 comments X

The pith

Multi-agent LLM deliberation functions as an input-dependent mixture of experts

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The authors apply the Friedkin-Johnsen model of opinion dynamics to multi-agent large language model systems. They demonstrate that the model's parameters for stubbornness and influence are not fixed but instead depend on the particular input being processed. This input dependence effectively makes the multi-agent system behave like a mixture of experts, where different agents exert more influence on different tasks. If true, this means such systems can achieve better performance than either a single agent or a static combination of agents, provided that the influence is allocated according to each agent's competence on the given input. The work then explores how observable factors such as an agent's self-assessed confidence and initial alignment with others determine who becomes influential.

Core claim

The Friedkin-Johnsen parameters governing multi-agent deliberation are input-dependent, which recasts multi-agent systems as mixtures of experts capable of outperforming single agents and static ensembles when the routing of influence aligns with agent competence.

What carries the argument

Friedkin-Johnsen opinion dynamics, a model that tracks how individual stubbornness and interpersonal influence shape opinion updates over time in a group.

Load-bearing premise

The Friedkin-Johnsen model accurately represents the deliberation and opinion change processes observed in groups of large language models.

What would settle it

A direct measurement showing that the effective influence exerted by each agent in a multi-agent LLM discussion does not change as a function of the input topic or question.

Figures

Figures reproduced from arXiv: 2605.25929 by Franka Bause, Jonas Niederle, Martin Pawelczyk, Rebekka Burkholz.

Figure 1
Figure 1. Figure 1: FJ induces MoE. (a) FJ parameter variability across MMLU-Pro questions for GPT-5.4 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Association of competence and confidence for ChatGPT-5.4 Mini on MMLU-Pro. Hard routing towards the most competent agent. As a special case, we can also consider hard routing, where a single agent j ′ (S) receives all weight, leading to the condition E hPn j=1 aj rj (S) − rj ′(S)(S) i > E[Da(S)], because the local diversity Dπ(S)(S) = 0. Thus, hard routing must compensate for the diversity benefit with a s… view at source ↗
Figure 3
Figure 3. Figure 3: Example: Initial beliefs of agents (left). FJ weight matrix, [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Influence (mean ± 95% confidence interval) for different communication styles, differences in influence suggest perceived confidence plays a role in FJ dynamics. We empirically study the FJ dynamics in agentic sys￾tems and investigate how the routing weights are dependent on the different variables. If not indicated otherwise, the plots show results from the MMLU￾pro dataset with communication style prompt… view at source ↗
Figure 5
Figure 5. Figure 5: Logistic regression coefficients clas￾sifying the most influential agent. Confidence and competence are the strongest positive pre￾dictors. Communication styles coefficients suggest perceived confidence plays a role in FJ dynamics. Confidence leads to influence. To understand the routing behavior, we investigate the relation between metrics of the initial beliefs and competence with in￾fluence of agents by… view at source ↗
Figure 6
Figure 6. Figure 6: Parameter variability over all samples no special prompts for [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Parameter variability over all samples no special prompts for [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Confidence vs. Influence (left) and confidence vs. peer influence (right) with no special [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Confidence vs. Influence (left) and confidence vs. peer influence (right) with no special [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Probability of changing answer to the majority answer for cases where the agent is [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Results for GPT-5.4 Mini on the MMLU-Pro dataset. Roles seem to have an even bigger influence on perceived confidence as seen by the weights agents receive from the others. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Results for Qwen2.5-14B-Instruct on the MMLU-Pro dataset. The same tendencies as for GPT-5.4 Mini can be seen. confidence competence student teacher mathematician doctor rookie Rel. Confidence Alignment Alignm. Score Alignm. Count 0.0 0.1 0.2 0.3 Gini Importance R 2 (train) = 0.399 ± 0.054 R 2 (CV) = 0.285 ± 0.032 (a) confidence competence student teacher mathematician doctor rookie Rel. Confidence Alignm… view at source ↗
Figure 13
Figure 13. Figure 13: Results for the MMLU-Pro dataset with GPT-5.4 Mini and communication style prompts. Coefficients for a random forest regression predicting influence (a) and a random forest classifying the most influential agent in a MAS (b), with reported CV R2 and Acc respectively. The high CV score, especially on the classification task (b) suggests the defined variables are highly predictive of the FJ dynamics. 23 [P… view at source ↗
Figure 14
Figure 14. Figure 14: Results for the MMLU-Pro dataset with GPT-5.4 Mini and communication style prompts. When relating confidence and relative confidence to competence (a), we see a clear trend: confidence agents are more competent. We report Spearman’s ranked correlation per MAS. (b) An agents influence is related to its stubbornness γ. 0.00 0.25 0.50 0.75 1.00 Influence 0.0 0.2 0.4 0.6 0.8 1.0 Belief in correct answer r per… view at source ↗
Figure 15
Figure 15. Figure 15: Results for the MMLU-Pro dataset with GPT-5.4 Mini and communication style prompts. More competent agents gain more influence (a) and relative influence (b), we again report Spearman’s ranked correlation per MAS. sample 186 sample 215 sample 223 sample 151 sample 113 sample 139 sample 193 sample 266 sample 288 sample 290 Init Final 0 1 W 0 1 gamma Indicator (intensity = conf.) Correct Incorrect [PITH_FUL… view at source ↗
Figure 16
Figure 16. Figure 16: Exemplary questions of MMLU-Pro with communication style prompts with [PITH_FULL_IMAGE:figures/full_fig_p024_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Confusion matrices showing empirical result of whether per sample version of Eq. [PITH_FULL_IMAGE:figures/full_fig_p025_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Weight heatmaps for samples from the MMLU-Pro dataset with [PITH_FULL_IMAGE:figures/full_fig_p030_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: More weight heatmaps for samples from the MMLU-Pro dataset with [PITH_FULL_IMAGE:figures/full_fig_p031_19.png] view at source ↗
read the original abstract

The effectiveness of multi-agent LLM deliberation depends not only on the agents' individual predictions, but also on how they communicate and collaborate. We study this mechanism through the lens of Friedkin-Johnsen (FJ) opinion dynamics, a tractable model for analyzing stubbornness, influence, and opinion change in multi-agent systems that captures empirically observed deliberation patterns. We show that the FJ parameters are input-dependent, turning multi-agent deliberation into a mixture of experts. This perspective implies that multi-agent systems can outperform single agents and static ensembles when routing reflects agent competence. Since competence is latent in practice, we analyze how influence is established through observable proxies: agents' self-assessed confidence, their perceived confidence, and initial alignment with other agents' views.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper models multi-agent LLM deliberation using the Friedkin-Johnsen (FJ) opinion dynamics framework. It claims that the FJ parameters are input-dependent, which recasts the deliberation process as a mixture of experts. This is used to argue that multi-agent systems can outperform single agents and static ensembles when influence routing aligns with latent agent competence, with influence analyzed via observable proxies including self-assessed confidence, perceived confidence, and initial alignment.

Significance. If the central modeling assumption holds, the work supplies a tractable analytic lens for influence and routing in multi-agent LLM systems, connecting classical opinion dynamics to modern LLM collaboration and offering a route to dynamic, competence-aware ensembles.

major comments (2)
  1. [Abstract] Abstract: The claim that the FJ model 'captures empirically observed deliberation patterns' is presented without any quantitative support (parameter recovery, trajectory matching, or predictive accuracy against observed LLM opinion updates). This assumption is load-bearing for the subsequent interpretation that input-dependent parameters create a mixture-of-experts routing mechanism.
  2. [Abstract] The manuscript provides no empirical validation that the FJ model accurately reproduces deliberation dynamics in actual multi-agent LLM interactions. Without such evidence, the claims that input-dependent parameters enable performance gains over static ensembles and that competence can be proxied by the listed observables do not transfer to the target systems.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and detailed report. The two major comments both concern the empirical grounding of the Friedkin-Johnsen modeling assumption. We address each point below and will revise the manuscript to clarify the scope of our claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that the FJ model 'captures empirically observed deliberation patterns' is presented without any quantitative support (parameter recovery, trajectory matching, or predictive accuracy against observed LLM opinion updates). This assumption is load-bearing for the subsequent interpretation that input-dependent parameters create a mixture-of-experts routing mechanism.

    Authors: We agree the abstract phrasing is imprecise. The statement references the established literature on FJ dynamics in social systems rather than new quantitative validation on LLM trajectories. We will revise the abstract to read that the FJ model 'provides a tractable analytic framework previously shown to capture patterns in human deliberation' and will add an explicit caveat that direct parameter recovery or trajectory matching on LLM data is left for future work. The mixture-of-experts interpretation follows mathematically from the demonstrated input dependence and does not require the stronger empirical claim. revision: yes

  2. Referee: [Abstract] The manuscript provides no empirical validation that the FJ model accurately reproduces deliberation dynamics in actual multi-agent LLM interactions. Without such evidence, the claims that input-dependent parameters enable performance gains over static ensembles and that competence can be proxied by the listed observables do not transfer to the target systems.

    Authors: The manuscript is primarily a theoretical contribution that derives the input-dependent FJ equivalence and the resulting mixture-of-experts view, then analyzes observable influence proxies under that model. We do not present new experiments showing that FJ reproduces LLM deliberation trajectories. We will therefore (i) add a limitations paragraph stating that performance gains are conditional on the modeling assumption holding for LLMs and (ii) rephrase the relevant claims to indicate they are implications of the model rather than validated predictions. These changes will prevent over-transfer of the results to deployed systems. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation relies on external model assumption without self-referential reduction

full rationale

The provided abstract and context present the Friedkin-Johnsen model as an adopted lens that 'captures empirically observed deliberation patterns,' followed by analysis showing input-dependent parameters that reframe deliberation as a mixture of experts. No equations, parameter-fitting steps, or self-citations are quoted that reduce any claimed prediction or uniqueness result to the inputs by construction. The central modeling choice functions as an independent assumption rather than a fitted quantity renamed as output, and the paper's claims about competence proxies and performance advantages remain logically downstream of that assumption without circular collapse. This is the most common honest finding for papers that introduce an external dynamical model.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract only; no free parameters, axioms, or invented entities are specified in the provided text.

pith-pipeline@v0.9.1-grok · 5660 in / 981 out tokens · 23170 ms · 2026-06-29T19:24:25.744083+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

39 extracted references · 5 canonical work pages · 1 internal anchor

  1. [1]

    Don’t trust stubborn neighbors: A security framework for agentic networks.CoRR, abs/2603.15809, 2026

    Samira Abedini, Sina Mavali, Lea Schönherr, Martin Pawelczyk, and Rebekka Burkholz. Don’t trust stubborn neighbors: A security framework for agentic networks.CoRR, abs/2603.15809, 2026

    work page arXiv 2026

  2. [2]

    Pan, Shuyi Yang, Lakshya A

    Mert Cemri, Melissa Z. Pan, Shuyi Yang, Lakshya A. Agrawal, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Dan Klein, Kannan Ramchandran, Matei Zaharia, Joseph E. Gonzalez, and Ion Stoica. Why do multi-agent LLM systems fail?, 2025

    2025

  3. [3]

    A survey on llm-based multi-agent system: Recent advances and new frontiers in application

    Shuaihang Chen, Yuanxing Liu, Wei Han, Weinan Zhang, and Ting Liu. A survey on llm- based multi-agent system: Recent advances and new frontiers in application.arXiv preprint arXiv:2412.17481, 2024

    work page arXiv 2024

  4. [4]

    Towards understanding mixture-of-experts layer in deep learning

    Zixiang Chen, Yihe Deng, Yue Wu, Quanquan Gu, and Yuanzhi Li. Towards understanding mixture-of-experts layer in deep learning. InAdvances in Neural Information Processing Systems, volume 35, 2022

    2022

  5. [5]

    Morris H. DeGroot. Reaching a consensus.Journal of the American Statistical Association, 69(345):118–121, 1974

    1974

  6. [6]

    Dietterich

    Thomas G. Dietterich. Ensemble methods in machine learning. InMultiple Classifier Systems, pages 1–15. Springer, 2000

    2000

  7. [7]

    Tenenbaum, and Igor Mordatch

    Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. Improving factuality and reasoning in language models through multiagent debate. InProceedings of the 41st International Conference on Machine Learning, volume 235 ofProceedings of Machine Learning Research, pages 11409–11431. PMLR, 2024

    2024

  8. [8]

    Friedkin and Eugene C

    Noah E. Friedkin and Eugene C. Johnsen. Social influence and opinions.Journal of Mathemati- cal Sociology, 15(3–4):193–206, 1990

    1990

  9. [9]

    Friedkin and Eugene C

    Noah E. Friedkin and Eugene C. Johnsen.Social Influence Network Theory: A Sociological Examination of Small Group Dynamics. Cambridge University Press, 2011

    2011

  10. [10]

    Opinion dynamics and mutual influence with llm agents through dialog simulation.arXiv preprint arXiv:2602.12583, 2026

    Yulong He, Dutao Zhang, Sergey Kovalchuk, Pengyi Li, and Artem Sedakov. Opinion dynamics and mutual influence with llm agents through dialog simulation.arXiv preprint arXiv:2602.12583, 2026

    work page arXiv 2026

  11. [11]

    Metagpt: Meta programming for a multi-agent collaborative framework

    Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber. Metagpt: Meta programming for a multi-agent collaborative framework. InInternational Conference on Learning Representations, 2024

    2024

  12. [12]

    Expert personas improve llm alignment but damage accuracy: Bootstrapping intent-based persona routing with prism, 2026

    Zizhao Hu, Mohammad Rostami, and Jesse Thomason. Expert personas improve llm alignment but damage accuracy: Bootstrapping intent-based persona routing with prism, 2026

    2026

  13. [13]

    Jacobs, Michael I

    Robert A. Jacobs, Michael I. Jordan, Steven J. Nowlan, and Geoffrey E. Hinton. Adaptive mixtures of local experts.Neural Computation, 3(1):79–87, 1991. 10

    1991

  14. [14]

    Wenxin Jiang and Martin A. Tanner. Hierarchical mixtures-of-experts for generalized linear models: Some results on denseness and consistency. InProceedings of the Seventh International Workshop on Artificial Intelligence and Statistics, volume R2 ofProceedings of Machine Learning Research, 03–06 Jan 1999

    1999

  15. [15]

    Jordan and Robert A

    Michael I. Jordan and Robert A. Jacobs. Hierarchical mixtures of experts and the em algorithm. Neural Computation, 6(2):181–214, 1994

    1994

  16. [16]

    Changgeon Ko, Jisu Shin, Hoyun Song, Huije Lee, Eui Jun Hwang, and Jong C. Park. Social dy- namics as critical vulnerabilities that undermine objective decision-making in LLM collectives, 2026

    2026

  17. [17]

    Better zero-shot reasoning with role-play prompting

    Aobo Kong, Shiwan Zhao, Hao Chen, Qicheng Li, Yong Qin, Ruiqi Sun, Xin Zhou, Enzhi Wang, and Xiaohang Dong. Better zero-shot reasoning with role-play prompting. InProceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), NAACL 2024, Mexico City, M...

    2024

  18. [18]

    Towards ethical multi-agent sys- tems of large language models: A mechanistic interpretability perspective.arXiv preprint arXiv:2512.04691, 2025

    Jae Hee Lee, Anne Lauscher, and Stefano V Albrecht. Towards ethical multi-agent sys- tems of large language models: A mechanistic interpretability perspective.arXiv preprint arXiv:2512.04691, 2025

    work page arXiv 2025

  19. [19]

    CAMEL: Communicative agents for ”mind” exploration of large language model society

    Guohao Li, Hasan Abed Al Kader Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. CAMEL: Communicative agents for ”mind” exploration of large language model society. InThirty-seventh Conference on Neural Information Processing Systems, 2023

    2023

  20. [20]

    Alicia Parrish, Angelica Chen, Nikita Nangia, Vishakh Padmakumar, Jason Phang, Jana Thomp- son, Phu Mon Htut, and Samuel R. Bowman. BBQ: A hand-built bias benchmark for question answering. In Smaranda Muresan, Preslav Nakov, and Aline Villavicencio, editors,Findings of the Association for Computational Linguistics: ACL 2022, Dublin, Ireland, May 22-27, 20...

    2022

  21. [21]

    Parsegov, Anton V

    Sergey E. Parsegov, Anton V . Proskurnikov, Roberto Tempo, and Noah E. Friedkin. Novel multidimensional models of opinion dynamics in social networks.IEEE Transactions on Automatic Control, 62(5):2270–2285, May 2017

    2017

  22. [22]

    Proskurnikov and Roberto Tempo

    Anton V . Proskurnikov and Roberto Tempo. A tutorial on modeling and analysis of dynamic social networks. part i.Annual Reviews in Control, 43:65–79, 2017

    2017

  23. [23]

    Chatdev: Communicative agents for software development

    Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen, Yusheng Su, Xin Cong, et al. Chatdev: Communicative agents for software development. InProceedings of the 62nd annual meeting of the association for computational linguistics (volume 1: Long papers), pages 15174–15186, 2024

    2024

  24. [24]

    When shift happens - confounding is to blame

    Abbavaram Gowtham Reddy, Celia Rubio-Madrigal, Rebekka Burkholz, and Krikamol Muandet. When shift happens - confounding is to blame. InThe Fourteenth International Conference on Learning Representations, 2026

    2026

  25. [25]

    Boosting for predictive sufficiency

    Abbavaram Gowtham Reddy, Rajeev Verma, Celia Rubio-Madrigal, Krikamol Muandet, and Rebekka Burkholz. Boosting for predictive sufficiency. InThe Fourteenth International Conference on Learning Representations, 2026

    2026

  26. [26]

    Explainability in human–agent systems.Autonomous agents and multi-agent systems, 33(6):673–705, 2019

    Avi Rosenfeld and Ariella Richardson. Explainability in human–agent systems.Autonomous agents and multi-agent systems, 33(6):673–705, 2019

    2019

  27. [27]

    Should we be going MAD? A look at multi-agent debate strategies for LLMs

    Andries Petrus Smit, Nathan Grinsztajn, Paul Duckworth, Thomas D Barrett, and Arnu Pretorius. Should we be going MAD? A look at multi-agent debate strategies for LLMs. In Ruslan Salakhutdinov, Zico Kolter, Katherine Heller, Adrian Weller, Nuria Oliver, Jonathan Scarlett, and Felix Berkenkamp, editors,Proceedings of the 41st International Conference on Mac...

    2024

  28. [28]

    Robustness of mixtures of experts to feature noise

    Dong Sun, Rahul Nittala, and Rebekka Burkholz. Robustness of mixtures of experts to feature noise. InForty-third International Conference on Machine Learning, 2026

    2026

  29. [29]

    Commonsenseqa: A question answering challenge targeting commonsense knowledge

    Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. Commonsenseqa: A question answering challenge targeting commonsense knowledge. InProceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, MN, USA, June 2-7, 2019, Volume 1 (...

    2019

  30. [30]

    Opinion dynamics in social networks with stubborn agents: An issue-based perspective.Automatica, 96:213–223, 2018

    Ye Tian and Long Wang. Opinion dynamics in social networks with stubborn agents: An issue-based perspective.Automatica, 96:213–223, 2018

    2018

  31. [31]

    Single-Agent LLMs Outperform Multi-Agent Systems on Multi-Hop Reasoning Under Equal Thinking Token Budgets

    Dat Tran and Douwe Kiela. Single-agent llms outperform multi-agent systems on multi-hop reasoning under equal thinking token budgets.arXiv preprint arXiv:2604.02460, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  32. [32]

    Decoding echo chambers: LLM-powered simulations revealing polarization in social networks

    Chenxi Wang, Zongfang Liu, Dequan Yang, and Xiuying Chen. Decoding echo chambers: LLM-powered simulations revealing polarization in social networks. In Owen Rambow, Leo Wanner, Marianna Apidianaki, Hend Al-Khalifa, Barbara Di Eugenio, and Steven Schockaert, editors,Proceedings of the 31st International Conference on Computational Linguistics, pages 3913–3...

    2025

  33. [33]

    Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou

    Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V Le, Ed H. Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. InThe Eleventh International Conference on Learning Representations, 2023

    2023

  34. [34]

    MMLU-pro: A more robust and challenging multi-task language understanding benchmark

    Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, Tianle Li, Max Ku, Kai Wang, Alex Zhuang, Rongqi Fan, Xiang Yue, and Wenhu Chen. MMLU-pro: A more robust and challenging multi-task language understanding benchmark. InThe Thirty-eight Conference on Neural Information Processin...

    2024

  35. [35]

    Chi, Quoc V

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V . Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models. In Sanmi Koyejo, S. Mohamed, A. Agarwal, Danielle Belgrave, K. Cho, and A. Oh, editors,Advances in Neural Information Processing Systems 35: Annual Conference on Neura...

    2022

  36. [36]

    John Wiley & Sons, 2 edition, 2009

    Michael Wooldridge.An Introduction to MultiAgent Systems. John Wiley & Sons, 2 edition, 2009

    2009

  37. [37]

    Opinion dynamics in learning systems, 2026

    Jiduan Wu, Rediet Abebe, and Celestine Mendler-Dünner. Opinion dynamics in learning systems, 2026

    2026

  38. [38]

    Autogen: Enabling next-gen LLM applications via multi-agent conversations

    Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, Ahmed Hassan Awadallah, Ryen W White, Doug Burger, and Chi Wang. Autogen: Enabling next-gen LLM applications via multi-agent conversations. InFirst Conference on Language Modeling, 2024

    2024

  39. [39]

    teacher": ###Role### You are an excellent teacher and always teach your students problems correctly

    Hangfan Zhang, Zhiyao Cui, Jianhao Chen, Xinrun Wang, Qiaosheng Zhang, Zhen Wang, Dinghao Wu, and Shuyue Hu. Stop overvaluing multi-agent debate – we must rethink evaluation and embrace model heterogeneity, 2025. 12 A Theory: Proofs of theorems and discussion Equilibrium beliefs are convex ensembles.The Friedkin-Johnsen dynamics induce a graph- dependent ...

    2025