This study contributes to the field of artificial intelligence by offering a structured approach to building, testing, and refining multi-agent architectures that balance knowledge grounding, perspective modelling, and reasoning validation.
Abstract
Large Language Models have rapidly advanced, enabling composite multi-agent systems capable of sophisticated text-based collaboration. In domains that demand coordinated reasoning, teams of agents can combine complementary strengths to produce better results than a single model. In practice, multi-agent decision making faces recurring shortcomings: agents frequently operate with poorly grounded or outdated knowledge, they seldom represent peers’ beliefs and intentions in a disciplined manner, and their conclusions are rarely checked against explicit logical constraints. These gaps reduce coherence, hinder dependable cooperation, and make the reasoning chain difficult to explore, often manifesting their black-box nature. How can a practical MAS architecture be designed to ground agents with relevant information, support explicit modelling of collaborators’ beliefs, and validate outputs against formal rules? How can the individual and synergized contributions of these mechanisms be measured across multiple case studies? To explore these questions, we present a purpose-built experimental platform. The system integrates a retrieval-augmented generation pipeline to ground agents in relevant materials, a Theory of Mind module that makes agent beliefs explicit, and an LLM-based logic auditor that evaluates outputs against an explicit rule set. Applied to diverse text-based decision-support and coordination case studies, the platform and protocol provide a reproducible blueprint for investigating how grounding, social reasoning, and logic validation together improve multi-agent collaboration. In doing so, this study contributes to the field of artificial intelligence by offering a structured approach to building, testing, and refining multi-agent architectures that balance knowledge grounding, perspective modelling, and reasoning validation.
This work introduces the Language Model Council (LMC), a collaborative framework that combines the expertise of multiple specialized AI agents to evaluate a user query from different perspectives and outperforms traditional single-model systems by improving response quality, reducing hallucinations, and increasing user trust through enhanced explainability.
D. M, Shwetha Kr, G. Divya et al.· International Research Journ...· 0 citations
Structured multi-agent debates among Large Language Models (LLMs) have emerged as a powerful paradigm for enhancing reasoning reliability and argumentative coherence. Motivated by the European Space Agency’s (ESA) interest in trustworthy AI for space operations, this study proposes a moderated, domain-adaptive multi-agent debate framework applied to the high-stakes domain of satellite communications (SatCom). Specifically, it assesses (i) the efficacy of structured deliberation against single-agent baselines, and (ii) the impact of model heterogeneity versus homogeneity. A single-agent baseline is compared against a multi-agent framework deploying a moderator and two domain-specialized experts. These systems utilize local 70B-parameter LLMs in homogeneous (Llama-3.3) and heterogeneous (Llama-3.3 + DeepSeek-R1 + Qwen-2.5) configurations, all augmented with a shared, curated Retrieval-Augmented Generation (RAG) corpus combining academic institutional sources and ESA material from the Nebula portal (SatNex V programme). Outputs from 213 technical queries are evaluated via LLM-as-a-judge across three phases: baseline proficiency, strategic reasoning, and executive readiness. Single-agent systems lead in encyclopedic tasks, where retrieval suffices over deliberation. However, both multi-agent configurations outperform in strategic reasoning, with heterogeneous debates achieving superior performance in executive scenarios by victory margins of up to 2.75 points on a 10-point scale. These results validate architectural diversity as a decisive factor in resolving high-complexity technical conflicts. Ultimately, this work delivers a generalizable, fully traceable deliberation framework suitable for real-world, mission-critical environments. Code, prompts, and evaluation data are publicly available at
https://github.com/amozo-es/multi-agent-debate/
.
Susana Gómez Álvarez, Alejandro Mozo Quesada, Tomás Navarro et al.· Journal of Intelligence and...· 0 citations
An organizing framework for understanding LLM‐based agents is established, systematically deconstructing both single‐agent and multi‐agent systems into their core components, and the architectural principles and key mechanisms that underpin their intelligence are analyzed.
Yuheng Cheng, Ceyao Zhang, Zheng-Wen Zhang et al.· WIREs Data Mining and Knowle...· 1 citation
Multi-Agent Debate (MAD) is a promising paradigm for improving the accuracy and robustness of Large Language Model (LLM)-based agentic systems. It enables multiple agents to exchange arguments, critique each other's outputs, and iteratively converge towards a solution. However, research remains fragmented, with inconsistent terminology and no rigorous synthesis of MAD design dimensions. We present a systematic literature review characterizing 141 primary studies on MAD. We derive a three-dimensional taxonomy covering debate participants, the interaction mechanisms structuring the exchange, and the agreement protocols governing debate resolution, supported by formal notations to render MAD configurations. Our analysis reveals that the field has implicitly converged on a narrow design pattern - static, fully connected topologies, verbatim exchange, short-term memory and voting resolution strategies - adopted by convention rather than systematic comparison, while promising alternatives remain marginal. Because any MAD setting reflects roughly a dozen interacting design decisions, cross-study comparison is unreliable when these are left implicit. We position the taxonomy as a descriptive map of the research landscape, a framework for controlled benchmarking, and potentially as a schema for machine-readable MAD specifications. As future work, we propose formalizing it into an executable specification, enabling cost-aware benchmarking and automated tuning of debate configurations.
Quim Motger, Marc Oriol, Jordi Marco et al.· 1 citation
With increasingly powerful large language models (LLMs) and LLM-based agents tackling an ever-growing list of tasks, we envision a future where numerous LLM agents work seamlessly with other AI agents and humans to solve complex problems and enhance daily life. To achieve these goals, LLM agents must develop collaborative skills such as effective persuasion, assertion and disagreement, which are often overlooked in the prevalent single-turn training and evaluation of LLMs. In this work, we present Collaborative Reasoner ( Coral ), a framework to evaluate and improve the collaborative reasoning abilities of language models. In particular, tasks and metrics in Coral necessitate agents to disagree with incorrect solutions, convince their partners of a correct solution, and ultimately agree as a team to commit to a final solution, all through a natural multi-turn conversation. Through comprehensive evaluation on six collaborative reasoning tasks covering domains of coding, math, scientific QA and social reasoning, we show that current models cannot effectively collaborate due to undesirable social behaviors, collapsing even on problems that they can solve singlehandedly. To improve the collaborative reasoning capabilities of LLMs, we propose a self-play method to generate synthetic multi-turn preference data and further train the language models to be better collaborators. Experiments with Llama-3.1 , Ministral and Qwen-2.5 models show that our proposed self-improvement approach consistently outperforms finetuned chain-of-thought performance of the same base model, yielding gains up to 16.7% absolute. Human evaluations show that the models exhibit more effective disagreement and produce more natural conversations after training on our synthetic interaction data. 1
Ansong Ni, Ruta Desai, Yang Li et al.· Advances in Neural Informati...· 7 citations
This work conducts a systematic study of overthinking and overretrieval in search agents from a scaling perspective and proposes a lightweight post-hoc reflection framework that converts the proposed evaluation signals into explicit feedback rewards to guide agents' reasoning trajectories.
Xin Liu, Ruqing Zhang, Yu-An Liu et al.· Annual International ACM SIG...· 0 citations