# iMAD: Intelligent Multi-Agent Debate for Efficient and Accurate LLM Inference

arXiv: https://arxiv.org/abs/2511.11306 · HTML: https://mlsanity.com/paper/2511.11306/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use.
- Builds on 24 of its 39 references that are in the map.
- First posted on arXiv 14 Nov 2025.

## Placed in

- LLM agents and tool use / [Social simulation and multi-agent debate](https://mlsanity.com/agents/social-simulation-and-multi-agent-debate/)

## Builds on (most cited first, 24 of 24)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [A Unified Approach to Interpreting Model Predictions](https://mlsanity.com/paper/1705.07874/) (2017)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation](https://mlsanity.com/paper/2308.08155/) (2023)
- [Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering](https://mlsanity.com/paper/2209.09513/) (2022)
- [What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams](https://mlsanity.com/paper/2009.13081/) (2020)
- [PubMedQA: A Dataset for Biomedical Research Question Answering](https://mlsanity.com/paper/1909.06146/) (2019)
- [OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge](https://mlsanity.com/paper/1906.00067/) (2019)
- [Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate](https://mlsanity.com/paper/2305.19118/) (2023)
- [Measuring Calibration in Deep Learning](https://mlsanity.com/paper/1904.01685/) (2019)
- [Mixture-of-Agents Enhances Large Language Model Capabilities](https://mlsanity.com/paper/2406.04692/) (2024)
- [Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration](https://mlsanity.com/paper/2307.05300/) (2023)
- [Escape Sky-high Cost: Early-stopping Self-Consistency for Multi-step Reasoning](https://mlsanity.com/paper/2401.10480/) (2024)
- [On Verbalized Confidence Scores for LLMs](https://mlsanity.com/paper/2412.14737/) (2024)
- [GroupDebate: Enhancing the Efficiency of Multi-Agent Debate Using Group Discussion](https://mlsanity.com/paper/2409.14051/) (2024)
- [Rethinking Mixture-of-Agents: Is Mixing Different Large Language Models Beneficial?](https://mlsanity.com/paper/2502.00674/) (2025)
- [Stop Overvaluing Multi-Agent Debate -- We Must Rethink Evaluation and Embrace Model Heterogeneity](https://mlsanity.com/paper/2502.08788/) (2025)
- [Debate Only When Necessary: Adaptive Multiagent Collaboration for Efficient LLM Reasoning](https://mlsanity.com/paper/2504.05047/) (2025)
- [Confidence Calibration and Rationalization for LLMs via Multi-Agent Deliberation](https://mlsanity.com/paper/2404.09127/) (2024)
- [The choice of scaling technique matters for classification performance](https://mlsanity.com/paper/2212.12343/) (2022)
- [Group Think: Multiple Concurrent Reasoning Agents Collaborating at Token Level Granularity](https://mlsanity.com/paper/2505.11107/) (2025)
- [Can LLMs Beat Humans in Debating? A Dynamic Multi-agent Framework for Competitive Debate](https://mlsanity.com/paper/2408.04472/) (2024)
- [The Readability of Tweets and their Geographic Correlation with Education](https://mlsanity.com/paper/1401.6058/) (2014)

## Built on by (most cited first, 10 of 10)

- [A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agents](https://mlsanity.com/paper/2602.06052/) (2026)
- [Courtroom-Style Multi-Agent Debate with Progressive RAG and Role-Switching for Controversial Claim Verification](https://mlsanity.com/paper/2603.28488/) (2026)
- [HCP-MAD:Heterogeneous Consensus-Progressive Reasoning for Efficient Multi-Agent Debate](https://mlsanity.com/paper/2604.09679/) (2026)
- [Thinking Like a Clinician: A Cognitive AI Agent for Clinical Diagnosis via Panoramic Profiling and Adversarial Debate](https://mlsanity.com/paper/2604.23605/) (2026)
- [What Do Agents Communicate? Characterizing Information Exchange in Multi-Agent Systems](https://mlsanity.com/paper/2605.20548/) (2026)
- [FaLCon: Facet-Anchored Retrieval with Late Consensus for Sim2Real Text-Based Person Anomaly Search](https://mlsanity.com/paper/2608.09474/) (2026)
- [The impact of multi-agent debate protocols on debate quality: a controlled case study](https://mlsanity.com/paper/2603.28813/) (2026)
- [Confidence Composition for Multiagent Language Model Systems](https://mlsanity.com/paper/2606.13591/) (2026)
- [CERA-MoA: Co-Evolving Routing Mechanisms with Continually Learning LLM Agents](https://mlsanity.com/paper/2609.18779/) (2026)
- [Measuring Collapse and Correction in Homogeneous-Panel LLM Debate](https://mlsanity.com/paper/2609.35279/) (2026)

## Abstract (arXiv)

Large Language Model (LLM) agent systems have advanced rapidly, driven by their strong generalization in zero-shot settings. To further enhance reasoning and accuracy on complex tasks, Multi-Agent Debate (MAD) has emerged as a promising framework that engages multiple LLM agents in structured debates to encourage diverse reasoning. However, triggering MAD for every query is inefficient, as it incurs substantial computational (token) cost and may even degrade accuracy by overturning correct single-agent answers. To address these limitations, we propose intelligent Multi-Agent Debate (iMAD), a token-efficient framework that selectively triggers MAD only when it is likely to be beneficial (i.e., correcting an initially wrong answer). To achieve this goal, iMAD learns generalizable model behaviors to make accurate debate decisions. Specifically, iMAD first prompts a single agent to produce a structured self-critique response, from which we extract 41 interpretable linguistic and semantic features capturing hesitation cues. Then, iMAD uses a lightweight debate-decision classifier, trained using our proposed FocusCal loss, to determine whether to trigger MAD, enabling robust debate decisions without test dataset-specific tuning. Through extensive experiments using six (visual) question answering datasets against five competitive baselines, we have shown that iMAD significantly reduces token usage (by up to 92%) while also improving final answer accuracy (by up to 13.5%).