mlsanityData to 7 Oct 2026

Beyond Majority Voting: LLM Aggregation by Leveraging Higher-Order Information

arXiv 2510.01499 ↗First posted 1 Oct 2025cs.LG cs.AI cs.GTcited by 25

On the map: LLM reasoning → Reasoning for agents, planning and code

also in 1 more
Abstract (arXiv)

With the rapid progress of multi-agent large language model (LLM) reasoning, how to effectively aggregate answers from multiple LLMs has emerged as a fundamental challenge. Standard majority voting treats all answers equally, failing to consider latent heterogeneity and correlation across models. In this work, we design two new aggregation algorithms called Optimal Weight (OW) and Inverse Surprising Popularity (ISP), leveraging both first-order and second-order information. Our theoretical analysis shows these methods provably mitigate inherent limitations of majority voting under mild assumptions, leading to more reliable collective decisions. We empirically validate our algorithms on synthetic datasets, popular LLM fine-tuning benchmarks such as UltraFeedback and MMLU, and a real-world healthcare setting ARMMAN. Our algorithms consistently outperform standard baselines, establishing a robust, training-free framework for effective multi-agent LLM aggregation.

Key facts. Cited by 25 papers in the map (data through 7 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 44 of its 68 references that are in the map. First posted on arXiv 1 Oct 2025.

Builds on

44 of 68 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Debate or Vote: Which Yields Better Decisions in Multi-Agent Large Language Models?24 Aug
  2. 2025Multiagent Finetuning: Self Improvement with Diverse Reasoning Chains10 Jan
  3. 2024Mixture-of-Agents Enhances Large Language Model Capabilities7 Jun
  4. 2023ReConcile: Round-Table Conference Improves Reasoning via Consensus among Diverse LLMs22 Sep

Other references · 40

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  4. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  5. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  6. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  7. 2024GPT-4o System Cardcited by 4,467
  8. 2024Qwen2 Technical Reportcited by 2,076
  9. 2024Mixtral of Expertscited by 1,785
  10. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  11. 2024Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preferencecited by 1,151
  12. 2024DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Modelcited by 1,093
28 more
  1. 2024DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Modelscited by 834
  2. 2024Phi-4 Technical Reportcited by 675
  3. 2023LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusioncited by 550
  4. 2024Debating with More Persuasive LLMs Leads to More Truthful Answerscited by 255
  5. 2025Deep Think with Confidencecited by 199
  6. 2023Universal Self-Consistency for Large Language Model Generationcited by 198
  7. 2024ForecastBench: A Dynamic Benchmark of AI Forecasting Capabilitiescited by 104
  8. 2025Harnessing Multiple Large Language Models: A Survey on LLM Ensemblecited by 86
  9. 2021Field Study in Deploying Restless Multi-Armed Bandits: Assisting Non-Profits in Improving Maternal and Child Healthcited by 43
  10. 2024Merge, Ensemble, and Cooperate! A Survey on Collaborative Strategies in the Era of Large Language Modelscited by 43
  11. 2024LLM-TOPLA: Efficient LLM Ensemble by Maximising Diversitycited by 42
  12. 2024LLM-Ensemble: Optimal Large Language Model Ensemble Method for E-commerce Product Attribute Value Extractioncited by 39
  13. 2024Serial Position Effects of Large Language Modelscited by 30
  14. 2024A Comparative Study on Annotation Quality of Crowdsourcing and LLM via Label Aggregationcited by 23
  15. 2024Reinforcement Learning from Multi-role Debates as Feedback for Bias Mitigation in LLMscited by 22
  16. 2024Language Model Council: Democratically Benchmarking Foundation Models on Highly Subjective Taskscited by 17
  17. 2021Are You Smarter Than a Random Expert? The Robust Aggregation of Substitutable Signalscited by 13
  18. 2014Elicitation for Aggregationcited by 9
  19. 2021The Wisdom of the Crowd and Higher-Order Beliefscited by 8
  20. 2019Forecast Aggregation via Peer Predictioncited by 7
  21. 2021Wisdom of the Crowd Voting: Truthful Aggregation of Voter Information and Preferencescited by 7
  22. 2021Eliciting Thinking Hierarchy without a Priorcited by 7
  23. 2023Robust Decision Aggregation with Second-order Informationcited by 7
  24. 2024Algorithmic Robust Forecast Aggregationcited by 5
  25. 2025Cost-aware LLM-based Online Dataset Annotationcited by 4
  26. 2023A Random Dictator Is All You Needcited by 1
  27. 2024Peer Expectation in Robust Forecast Aggregation: The Possibility/Impossibilitycited by 1
  28. 2025LLM-based Agent Simulation for Maternal Health Interventions: Uncertainty Estimation and Decision-focused Evaluationcited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 25.

25 papers.

Sort:

By area:

  1. 2026Don't Always Pick the Highest-Performing Model: An Information Theoretic View of LLM Ensemble Selectioncited by 10
  2. 2026Are Diversity Metrics Measuring Diversity? A Capability-Controlled Audit of Majority-Vote Gain in LLM Ensemblescited by 7
  3. 2026Epistemic Gain, Aleatoric Cost: Uncertainty Decomposition in Multi-Agent Debate for Math Reasoningcited by 5
  4. 2026Minority Sentinel: When to Overturn Majority Voting in Multi-Agent LLM Debatescited by 4
  5. 2026RAL2M: Retrieval Augmented Learning-To-Match Against Hallucination in Compliance-Guaranteed Service Systemscited by 2
  6. 2026Unifying Temporal and Structural Credit Assignment in LLM-Based Multi-Agent Prompt Optimizationcited by 2
  7. 2026Dependence-Aware Label Aggregation for LLM-as-a-Judge via Ising Modelscited by 1
  8. 2026Multi-LLM Query Optimizationcited by 1
  9. 2026Prior-Agnostic Robust Forecast Aggregationcited by 1
  10. 2026When Does Delegation Beat Majority? A Delegation-Based Aggregator for Multi-Sample LLM Inferencecited by 1
  11. 2026Robust Persona-Aware Toxicity Detection with Prompt Optimization and Learned Ensemblingcited by 0
  12. 2026SCoOP: Semantic Consistent Opinion Pooling for Uncertainty Quantification in Multiple Vision-Language Model Systemscited by 0
13 more
  1. 2026Multi-agent decision making: A Blackwell's informativeness approachcited by 0
  2. 2026Certifiable Semantic Agreement Among LLM Agents: What the Admissibility Instrument Decidescited by 0
  3. 2026Uncertainty-Aware Trust Estimation for Multi-LLM Systems via Structured Expert Judgementcited by 0
  4. 2026Evaluating Large Language Models for Symbolic Security Protocol Analysiscited by 0
  5. 2026Test-Time Augmentation for LLMs: When Input Diversity Beats Output Diversity at Matched Computecited by 0
  6. 2026Divergent-Convergent Reasoning: Scaling Test-Time Compute through Structured Solution Synthesiscited by 0
  7. 2026Diverse by Reasoning: Harnessing the Wisdom of LLM Crowds for Future Predictioncited by 0
  8. 2026Performance Foundations of Parallel & Distributed Reasoning Language Modelscited by 0
  9. 2026CUSP: Decomposable Collective Uncertainty for Multi-Agent Multimodal Reasoningcited by 0
  10. 2026When Agents Disagree: Bayesian Backward Reasoning as a Label-Free Anchor for Multi-Agent Collective Decision-Makingcited by 0
  11. 2026Count Evidence, Not Sentences: Tempered Evidence Fusion of LLM Judgments for Long-Text Value Measurementcited by 0
  12. 2026Social Circuits behind Multi-agent Echo Chamberscited by 0
  13. 2026Refinement as a Service: Algorithmic Predictor Refinementcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Beyond the Strongest LLM: Multi-Turn Multi-Agent Orchestration vs. Single LLMs on Benchmarks
  2. 2024An Electoral Approach to Diversify LLM-based Multi-Agent Collective Decision-Making
  3. 2026Beyond Majority Voting: Efficient Best-Of-N with Radial Consensus Score
  4. 2024LLM Voting: Human Choices and AI Collective Decision Making
  5. 2025Mitigating Judgment Preference Bias in Large Language Models through Group-Based Polling
  6. 2025The Majority is not always right: RL training for solution aggregation
  7. 2025Multi-Agent Debate for LLM Judges with Adaptive Stability Detection
  8. 2026Demystifying Multi-Agent Debate: The Role of Confidence and Diversity