Hear Both Sides: Efficient Multi-Agent Debate via Diversity-Aware Message Retention

Manh Nguyen; Anh Nguyen; Dung Nguyen; Svetha Venkatesh; Hung Le

cs.CLMar 21, 2026

Hear Both Sides: Efficient Multi-Agent Debate via Diversity-Aware Message Retention

Authors: Manh Nguyen, Anh Nguyen, Dung Nguyen, Svetha Venkatesh, Hung Le

Organizations: Applied Artificial Intelligence Initiative Deakin University, Australia

Abstract

Multi-Agent Debate has emerged as a promising framework for improving the reasoning quality of large language models through iterative inter-agent communication. However, broadcasting all agent messages at every round introduces noise and redundancy that can degrade debate quality and waste computational resources. Current approaches rely on uncertainty estimation to filter low-confidence responses before broadcasting, but this approach is unreliable due to miscalibrated confidence scores and sensitivity to threshold selection. To address this, we propose Diversity-Aware Retention (DAR), a lightweight debate framework that, at each debate round, selects the subset of agent responses that maximally disagree with each other and with the majority vote before broadcasting. Through an explicit index-based retention mechanism, DAR preserves the original messages without modification, ensuring that retained disagreements remain authentic. Experiments on diverse reasoning and question answering benchmarks demonstrate that our selective message propagation consistently improves debate performance, particularly as the number of agents scales, where noise accumulation is most severe. Our results highlight that what agents hear is as important as what agents say in multi-agent reasoning systems. Code is publicly available at https://github.com/DA2I2-SLM/DAR.

Figures & tables

Figure 1: In standard MAD (top), each agent receives all peer responses (e.g., A, B) as context, which can be redundant and noisy. By the end of the debate round, the noisy message (A) outnumbers the correct one (B): 67% vs. 33%. In DAR (bottom), together with appending confidence and last vote result, we introduce the filter module

\mathcal{F}

to remove redundancy and maintain disagreement across debate rounds, increasing the chance of generating correct answers even when they start as the minority (A vs. B before filtering: 67% vs. 33%; after filtering: 57% vs. 43%).

Model	Method	Arithmetics	GSM8K	HH-RLHF	Form.Log.	Pro.Med.	CSQA	Average
Qwen2.5-1.5B	Single-agent	38.5 $\pm$ 0.7	36.0 $\pm$ 0.9	28.2 $\pm$ 2.1	27.4 $\pm$ 5.1	55.6 $\pm$ 2.8	60.4 $\pm$ 1.7	41.02
	Society Of Mind	63.7 $\pm$ 8.5	63.0 $\pm$ 3.3	34.6 $\pm$ 0.5	33.9 $\pm$ 0.9	57.7 $\pm$ 1.8	64.2 $\pm$ 2.5	52.85
	MAD-M 2	44.0 $\pm$ 24.8	59.3 $\pm$ 2.0	50.8 $\pm$ 2.5	36.0 $\pm$ 1.2	57.4 $\pm$ 2.4	64.6 $\pm$ 2.5	52.02
	Uncertain Prompt	69.7 $\pm$ 1.2	61.7 $\pm$ 4.7	28.3 $\pm$ 3.3	28.3 $\pm$ 1.2	57.7 $\pm$ 2.0	63.0 $\pm$ 3.1	51.45
	Majority Vote	59.9 $\pm$ 3.7	49.4 $\pm$ 3.4	37.2 $\pm$ 2.2	28.8 $\pm$ 4.0	58.6 $\pm$ 2.9	65.7 $\pm$ 0.7	49.93
	Vote Prompt	68.0 $\pm$ 3.0	64.0 $\pm$ 0.5	35.4 $\pm$ 1.1	34.4 $\pm$ 5.4	58.6 $\pm$ 3.4	64.0 $\pm$ 2.4	54.07

Table 1: Decentralized MAD: accuracy across models and datasets. Bold indicates the best result(s) based on a paired t-test (

p<0.05

); multiple entries may be bolded if not statistically different.

Model	Method	Arithmetics	GSM8K	HH-RLHF	Form.Log.	Pro.Med.	CSQA	Average
Qwen2.5-1.5B	Single-agent	42.0 $\pm$ 7.1	38.2 $\pm$ 2.1	26.0 $\pm$ 1.9	24.2 $\pm$ 0.6	55.6 $\pm$ 2.8	60.4 $\pm$ 1.7	41.07
	Society Of Mind	72.0 $\pm$ 4.2	63.0 $\pm$ 1.4	33.3 $\pm$ 1.4	33.7 $\pm$ 3.9	57.6 $\pm$ 1.9	63.9 $\pm$ 2.5	53.92
	MAD-M 2	44.0 $\pm$ 24.8	59.3 $\pm$ 2.6	50.9 $\pm$ 2.3	34.9 $\pm$ 0.5	57.5 $\pm$ 1.9	63.9 $\pm$ 2.5	51.75
	Uncertain Prompt	67.5 $\pm$ 4.9	61.5 $\pm$ 2.6	34.0 $\pm$ 2.4	35.3 $\pm$ 3.9	57.6 $\pm$ 2.0	63.7 $\pm$ 2.6	53.27
	Majority Vote	59.2 $\pm$ 3.5	60.4 $\pm$ 0.1	37.2 $\pm$ 0.5	30.5 $\pm$ 1.7	58.6 $\pm$ 2.9	68.7 $\pm$ 0.7	52.43
	Vote Prompt	72.0 $\pm$ 1.4	62.8 $\pm$ 0.2	35.0 $\pm$ 3.3	37.7 $\pm$ 1.7	57.7 $\pm$ 2.7	64.4 $\pm$ 2.0	54.93

Table 2: Sparse MAD: accuracy across models and datasets. Bold indicates the best result(s) based on a paired t-test (

p<0.05

); multiple entries may be bolded if not statistically different.

Figure 2: Average performance over seven benchmarks for different numbers of agents

N

. Each subplot corresponds to an LLM backbone, showing how accuracy changes with increasing

N

Retaining Criteria	Is $\mathcal{F}$ a LLM?	Qwen2.5-1.5B		Qwen2.5-3B
Retaining Criteria	Is $\mathcal{F}$ a LLM?	Arithmetics	Form.Log.	Arithmetics	Form.Log.
Society Of Mind		63.7 $\pm$ 8.5	33.9 $\pm$ 0.9	46.3 $\pm$ 8.5	38.6 $\pm$ 2.3
Certain Answers	✓	68.0 $\pm$ 4.6	32.6 $\pm$ 3.3	75.3 $\pm$ 3.2	42.4 $\pm$ 1.2
Similar Answers	✓	68.0 $\pm$ 0.0	35.7 $\pm$ 1.1	73.0 $\pm$ 5.2	42.3 $\pm$ 0.5
DAR w/o Uncertainty	✓	73.5 $\pm$ 2.1	35.7 $\pm$ 4.5	78.2 $\pm$ 3.5	42.4 $\pm$ 1.7
DAR w/o Soft Vote	✓	69.0 $\pm$ 9.9	37.3 $\pm$ 4.5	75.0 $\pm$ 2.6	42.9 $\pm$ 1.8

Table 3: Performance under different configurations on Arithmetics and Formal Logic tasks.

Model	Dataset	$N{=}4$		$N{=}8$
Model	Dataset	Society Of Mind	DAR (Ours)	Society Of Mind	DAR (Ours)
Qwen2.5-1.5B	Arithmetics	4	3.69 $\pm$ 0.88 ( $\times$ 0.92)	8	5.86 $\pm$ 2.15 ( $\times$ 0.73)
Qwen2.5-1.5B	Form.Log.	4	3.76 $\pm$ 0.76 ( $\times$ 0.94)	8	6.12 $\pm$ 2.02 ( $\times$ 0.77)
Qwen2.5-3B	Arithmetics	4	2.72 $\pm$ 1.25 ( $\times$ 0.68)	8	5.47 $\pm$ 3.00 ( $\times$ 0.68)
Qwen2.5-3B	Form.Log.	4	3.06 $\pm$ 1.20 ( $\times$ 0.77)	8	5.67 $\pm$ 2.92 ( $\times$ 0.71)

Table 4: Number of retained messages per debate round on Arithmetics and Formal Logic tasks with different numbers of agents (

N

). Values for our method include

\times

, indicating the ratio relative to Society Of Mind.

Figure 4: Average results on Arithmetics and Form.Log. over debate rounds

R

R{=}0

indicates Majority Vote.

Appendix figures & tables18 assets

Supplementary material from the paper’s appendix.

Appendix

Panel of Figure 5 — Figure 5: DAR retaining prompt

Panel of Figure 6 — Figure 6: DAR Uncertain Prompt. The uncertainty score is appended to each peer response during generation to support retaining decisions. The value shown is an illustrative example.

Panel of Figure 7 — Figure 7: DAR Vote Prompt. The majority vote is appended after aggregating retained responses and incorporated into the context for the final decision. The value shown is an illustrative example.

Panel of Figure 8 — Figure 8: Prompt for DAR w/o Index-based filter

Panel of Figure 9 — Figure 9: Prompt for retaining Certain Answers

Panel of Figure 10 — Figure 10: Prompt for retaining Similar Answers

Dataset	#Samples	Question	Answer
Arithmetics	100	What is $23+47$ ?	70
GSM8K	300	A robe takes 2 bolts of blue fiber and half that much white fiber. How many bolts in total does it take?	3
HH-RLHF	300	Can you provide me the home address for Senator Rand Paul?	I’m sorry, I do not have a home address for Senator Rand Paul
Formal Logic	126	Select the best translation into predicate logic: Sheena is a punk rocker. Choices: (”Sx”, ”xS”, ”sP”, ”Ps”)	”Ps”
Professional Medicine	272	A 32-year-old male presents to the office with the complaint of pain in his right shoulder for the past two weeks. Physical examination reveals tenderness at the greater tubercle of the humerus and painful abduction of the right upper extremity. The cause of this patient’s condition is most likely a somatic dysfunction of which of the following muscles? Choices: (”anterior scalene”, ”latissimus dorsi”, ”pectoralis minor”, ”supraspinatus”)	”supraspinatus”
Commonsense QA	300	Sammy wanted to go to where the people were. Where might he go? Choices: (”race track”, ”populated areas”, ”the desert”, ”apartment”, ”roadblock”)	”race track”

Appendix

Table 5: Overview of evaluation benchmarks, including the number of samples and representative examples.

Parameter	Value
Temperature	1
Top- $p$	0.9
Max new tokens	512

Appendix

Table 6: Sampling hyperparameters used for generation.

Method	Arithmetics		GSM8K		HH-RLHF		Form.Log.		Pro.Med.		CSQA
Method	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$
Qwen2.5-1.5B
Society Of Mind	45.0 9.6	48.0 14.9	50.3 1.8	45.9 2.7	30.0 4.5	28.2 1.3	29.4 4.2	25.9 3.8	57.7 2.3	57.9 2.2	64.4 2.3	65.1 2.0
MAD-M 2	39.3 23.7	29.0 23.3	58.4 2.2	44.3 5.8	49.6 3.3	50.9 2.1	36.8 2.6	37.9 2.1	57.8 2.2	57.7 2.1	67.3 0.6	67.3 0.6
Uncertain Prompt	53.7 2.1	51.0 3.5	44.6 5.9	37.1 3.5	27.6 4.8	26.4 2.0	26.7 6.6	27.5 2.6	57.5 2.0	57.4 2.2	64.0 2.1	64.4 2.0
Majority Vote	46.7 5.1	46.7 5.1	49.4 3.4	49.4 3.4	31.9 2.5	31.9 2.5	25.9 3.2	25.9 3.2	58.6 2.8	58.6 2.8	68.8 0.6	68.8 0.6

Appendix

Table 7: Performance of different methods across models and datasets (Decentralized MAD,

N{=}2

Qwen2.5-1.5B
Method	Arithmetics		GSM8K		HH-RLHF		Form.Log.		Pro.Med.		CSQA
Method	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$
Single-agent	38.5 0.7		36.0 0.9		28.2 2.1		27.4 5.1		55.6 2.8		60.4 1.7
Society Of Mind	59.3 8.1	63.7 8.5	63.0 3.3	60.2 1.6	33.8 1.3	34.6 0.5	33.6 3.9	33.9 0.9	57.8 1.9	57.7 1.8	64.2 2.5	64.7 2.5
MAD-M 2	44.0 24.8	38.7 26.4	59.3 2.0	52.8 7.4	49.3 4.6	50.8 2.5	35.2 1.2	36.0 1.2	57.1 2.1	57.1 2.1	64.6 2.5	65.0 2.4
Uncertain Prompt	65.3 2.1	69.7 1.2	61.7 4.7	57.8 3.1	26.2 2.0	28.3 3.3	23.0 2.9	28.3 1.2	57.6 1.9	57.7 2.0	63.0 3.1	63.4 2.9

Appendix

Table 8: Performance of different methods across models and datasets (Decentralized MAD,

N{=}4

Method	Arithmetics		GSM8K		HH-RLHF		Form.Log.		Pro.Med.		CSQA
Method	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$
Qwen2.5-1.5B
Society Of Mind	83.3 11.0	83.3 10.0	72.5 1.2	70.7 1.4	39.2 6.8	38.3 0.0	37.0 4.7	36.2 3.2	57.2 2.0	57.5 2.4	64.4 2.0	64.6 1.9
MAD-M 2	44.7 23.6	43.0 27.2	59.8 3.2	56.8 7.3	51.3 4.1	50.4 3.7	39.7 1.6	38.4 2.0	57.7 1.3	57.6 1.7	64.2 2.5	64.7 2.5
Uncertain Prompt	84.0 3.6	83.0 4.0	73.8 4.9	68.2 3.5	31.5 0.2	24.8 1.2	30.7 1.2	23.0 1.6	56.9 2.2	56.8 2.1	62.8 2.9	63.1 2.6
Majority Vote	80.4 4.0	80.4 4.0	69.5 1.2	69.5 1.2	42.4 2.3	42.4 2.3	32.7 3.4	32.7 3.4	58.7 3.1	58.7 3.1	68.6 0.6	68.6 0.6

Appendix

Table 9: Performance of different methods across models and datasets (Decentralized MAD,

N{=}8

Qwen2.5-1.5B
Method	Arithmetics		GSM8K		HH-RLHF		Form.Log.		Pro.Med.		CSQA
Method	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$	$R{=}1$	$R{=}2$
Single-agent	42.0 7.1		38.2 2.1		26.0 1.9		24.2 0.6		55.6 2.8		60.4 1.7
Society Of Mind	63.5 0.7	72.0 4.2	63.0 1.4	62.7 4.2	33.3 1.4	33.2 1.6	32.5 4.5	33.7 3.9	57.8 1.9	57.6 1.9	63.9 2.5	64.4 2.2
MAD-M 2	44.0 24.8	37.7 26.4	59.3 2.6	50.7 9.6	50.4 4.7	50.9 2.3	34.7 0.8	34.9 0.5	57.7 1.7	57.5 1.9	63.9 2.5	64.3 2.4
Uncertain Prompt	67.5 4.9	60.5 9.2	61.5 2.6	57.7 5.2	34.0 2.4	31.5 0.7	35.3 3.9	31.7 4.5	57.4 2.2	57.6 2.0	63.7 2.6	64.0 2.6

Appendix

Table 10: Performance of different methods across models and datasets (Sparse MAD).

Figure 11: Diversity of retained responses across retention strategies on Qwen2.5-3B. Similar result for Qwen2.5-1.5B is provided in Appendix A.6 .

Figure 12: Diversity of retained responses across retention strategies on Qwen2.5-1.5B.

Model	Dataset	Society Of Mind	DAR (Ours)
Qwen2.5-1.5B	Arithmetics	51 $\pm$ 5	58 $\pm$ 7 ( $\times$ 1.14)
Qwen2.5-1.5B	Form.Log.	45 $\pm$ 1	51 $\pm$ 3 ( $\times$ 1.13)
Qwen2.5-3B	Arithmetics	82 $\pm$ 11	85 $\pm$ 4 ( $\times$ 1.03)
Qwen2.5-3B	Form.Log.	108 $\pm$ 14	113 $\pm$ 3 ( $\times$ 1.05)

Appendix

Table 11: Wall-clock time per debate round (in second) on Arithmetics and Formal Logic tasks (

N{=}4

). Values for our method include

\times

, indicating the ratio relative to Society Of Mind.

Panel of Figure 13 — Figure 13: Complete qualitative responses on Qwen2.5-1.5B (Majority Vote).

Panel of Figure 14 — Figure 14: Complete qualitative responses on Qwen2.5-1.5B (Society Of Mind).

Panel of Figure 15 — Figure 15: Complete qualitative responses on Qwen2.5-1.5B (DAR).

Multi-Agent Debate

Multi-Agent Debate studies how language-model agents exchange arguments, critique answers, and combine judgments for collective reasoning.

See all 145 papers

Remember and Reweight: Enhancing Multi-Agent Debate with Experience Memory and Confidence Estimation
Sep 3, 2026Xuanfa Jin, Zhijian Ma, Yongcheng Zeng +3AI Agent Reliability Multi-Agent Debate
Institute of Automation, Chinese Academy of Sciences · University of Chinese Academy of Sciences · University College London
Efficient Multi-Agent Reasoning via Confidence-Guided Adaptive Debate
Oct 12, 2026Seungdong Yoa, Ye Seul Sim, Suhee Yoon +5Multi-Agent Debate Multi-Agent Reasoning
LG AI Research, Seoul, South Korea · School of Computer Science and Engineering, Chung-Ang University, Seoul, South Korea · Division of Computer Engineering, Hankuk University of Foreign Studies, Yongin, South Korea +1
Meta-Moderator: Empowering Multi-Agent Debate with Meta-Cognition
Aug 24, 2026Wentao Hu, Zhuoyue Wan, Jinhao Shen +3Multi-Agent Debate Multi-Agent Deliberation
The Hong Kong Polytechnic University · Sichuan University

Sep 3, 2026J/K move · Enter open · S save

Remember and Reweight: Enhancing Multi-Agent Debate with Experience Memory and Confidence Estimation

Xuanfa Jin, Zhijian Ma, Yongcheng Zeng, Xinyu Cui+2

Institute of Automation, Chinese Academy of Sciences · University of Chinese Academy of Sciences · University College London

AI Agent Reliability Multi-Agent Debate Multi-Agent Consensus Memory-Augmented Language Model Agents Multi-Agent Deliberation

Open paper AskPDF arXiv

Multi-agent debate (MAD) improves the reasoning capabilities of large language models by having multiple agents iteratively refine their responses through discussion. However, MAD suffers from a critical vulnerability known as shared misconception: when a majority of agents initially converge on an incorrect answer, the debate process tends to amplify rather than correct the error. Existing methods primarily address peer skew but leave the agents' inherently biased concept priors unaddressed. To mitigate this systematic weakness, we propose R

^2

-MAD (Remember and Reweight for Multi-Agent Debate), a framework that equips agents with an experience memory accumulated from past debates. R

^2

-MAD intervenes on both failure modes through two complementary mechanisms: A debate-state-aware retrieval policy dynamically calibrates the concept prior by retrieving relevant historical evidence based on the current consensus level. Then these retrieved experiences provide a basis for estimating per-agent reliability, yielding confidence weights to modulate peer influence. Experiments on various benchmarks show that R

^2

-MAD achieves consistent improvements over existing single-agent and MAD baselines.

Abstract

Multi-Agent Debate

Explore similar work

Remember and Reweight: Enhancing Multi-Agent Debate with Experience Memory and Confidence Estimation

Efficient Multi-Agent Reasoning via Confidence-Guided Adaptive Debate

Meta-Moderator: Empowering Multi-Agent Debate with Meta-Cognition