The DOI matches its Crossref record; 21 of the 21 references it cites were found in a registry or an index.
Camel: Communicative agents for" mind" exploration of large language model society
Open accessAbstract
The rapid advancement of chat-based language models has led to remarkable progress in complex task-solving. However, their success heavily relies on human input to guide the conversation, which can be challenging and time-consuming. This paper explores the potential of building scalable techniques to facilitate autonomous cooperation among communicative agents, and provides insight into their "cognitive" processes. To address the challenges of achieving autonomous cooperation, we propose a novel communicative agent framework named role-playing. Our approach involves using inception prompting to guide chat agents toward task completion while maintaining consistency with human intentions. We showcase how role-playing can be used to generate conversational data for studying the behaviors and capabilities of a society of agents, providing a valuable resource for investigating conversational language models. In particular, we conduct comprehensive studies on instruction-following cooperation in multi-agent settings. Our contributions include introducing a novel communicative agent framework, offering a scalable approach for studying the cooperative behaviors and capabilities of multi-agent systems, and open-sourcing our library to support research on communicative agents and beyond: https://github.com/camel-ai/camel.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“The development of medical AI agents can be approached through two primary methodologies: graphical user interface (GUI)-based frameworks and coding-based development.110,111,112”
From A foundational architecture for AI agents in healthcare (2025) · cites this paper as [111]
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Cited by
Items 201–300 of 368
A survey of large language models: techniques, applications, and challenges
Ji, J., Zhao, X., & Shi, X. (2026) · Multimedia Systems · doi:10.1007/s00530-026-02288-9
“Error propagation in multi-step tasks presents a fundamental challenge, where initial mistakes compound through subsequent actions with amplifying effects [ 103 ].”
PaCR: An Agentic Preference-Aligned TAP Conflict Handling
Li, R., Liang, Z., Xiao, Y., Ye, Q., & Hu, H. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3693064
“In the context of smart-home service provision, we also observe the rise of LLM-powered agents [44], [45], [46] that autonomously perform tasks through iterative reasoning and tool usage.”
Multi-Agent LLM Framework for Imageability Assessment in Trusted Multimodal Data Spaces
Zhang, B., Huang, G., & Xu, C. (2025) · International Conference on Big Data Economy and Information Management (BDEIM) · doi:10.1145/3800000.3800222 · Cited by 1
“Recent optimization work includes Li et al. [17] introducing role-playing frameworks for agent coordination, Liang et al. [18] demonstrating that structured argumentation improves reasoning quality and reduces hallucinations, and Wu et al. [32] developing AutoGen for automated agent generation and…”
Agent Evolution: Capability-Mechanism Co-evolution
Chen, Y., Zhang, C., Zheng, X., & Tang, S. (2026) · International Conference on Generative Artificial Intelligence and Information Security (GAIIS) · doi:10.1109/gaiis69281.2026.11519194
“CAMEL and ChatDev by Li et al. and Qian et al. introduce “debate-review” mechanisms based on game theory, forcing multi-role cross-examination through non-adversarial game matrices, leveraging collective intelligence to correct individual blind spots and improving code first-pass rates [25], [26].”
Formal Model of a Controllable LLM Agent Based on a Cognitive Architecture
Dolgih, A. A. (2026) · Программные системы и вычислительные методы · doi:10.7256/2454-0714.2026.2.79851 · Cited by 1
Toward Human-Aligned, Modular, and Emotion-Aware Multi-Agent Systems: A Systematic Literature ReviewPreprint
Ghanem, H., Cruz, C., & Cherifi, H. (2026) · Research Square · doi:10.21203/rs.3.rs-9485421/v1
Czejdo, B., Daszczuk, W. B., & Grabski, W. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-27927-9_7
Ao, N., Li, B., Tian, W., Wang, F., & Wu, T. (2026) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2026.3704107
“Following the MAS catalog from MASLab [13], we compare with CoT [14], self-consistency [15], CAMEL [16], AutoGen [17], AgentVerse [18], DyLAN [19], LLM-Debate [20], MAV [21], EvoMAC [22], and MACNet [23].”
Divya, T., Prashanthi, V., Srikanth, V., & Venu Gopalachari, M. (2026) · International Conference on Inventive Research in Computing Applications (ICIRCA) · doi:10.1109/icirca69024.2026.11570324
“In MAS, multiple (typically role-specialized) agents work together to address issues that exceed the capacity of a single model [3], [19] by distributing tasks among agents [8], [13], [18].”
InfluentialMusicTutor: Facilitating Goal-Oriented Singing Practice via Multi-agent Tutoring Framework
Cheng, T., Li, X., Guo, T., Hou, M., Liu, Z., & Luo, W. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-29755-6_25
Adaptive PII Detection via Evolutionary Orchestration and Multi-agent Debate
Bang, G., Kim, D., Kim, D., & Min, M. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-1947-6_42
Fat-Cat: Document-Driven Metacognitive Multi-Agent System for Complex ReasoningPreprint
Yang, T., Wu, A., Wang, Y., Kang, J., Li, Y., Ren, J., Xin, Y., & Yuan, Y. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6990087
A Systematic Evaluation of Decision Mechanisms in Multi-agent LLMs for Misinformation Detection
Lee, Y., Huang, Y., & Chu, T.-H. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-30044-7_6
Bradford, M., & Krishnaswamy, N. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-29794-5_57
MAMA-V: A Verifiable Distributed Multi-Agent Model AI Framework for Secure Orchestration
Liu, S., Chalkiadakis, G. N., & Polyzos, G. C. (2026) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom59046.2026.11571481 · Cited by 1
“In open environments or Large Language Model (LLM) societies [3]–[5], this honesty assumption often fails.”
A Large Language Model-Enabled Multi-Agent Collaboration Method for Complex Task Solving
Wang, S., Feng, Y., & Fang, X. (2026) · International Symposium on Computer Technology and Information Science (ISCTIS) · doi:10.1109/isctis70043.2026.11572301
“Representative frameworks include CAMEL [11], MetaGPT [12], AgentVerse [13], AutoGen [14], and ChatDev [15], all of which demonstrate that communication protocols, role constraints, and modular coordination can increase solution quality on open-ended problems.”
Adaptive Goal Negotiation Network : From Role Emergence to Execution Governance in Multi-Agent LLMs
Tandon, S., Gupta, T., & Vijayalakshmi, M. (2026) · International Conference on Computing, Communication and Green Engineering (CCGE) · doi:10.1109/ccge67142.2026.11581522
“CAMEL and AgentVerse abandon predetermined roles and instead, allow coordination to arise because of interaction [7], [8].”
Jafaraghaeian, D., & Butvilas, T. (2026) · Applied Business Issues & Solutions · doi:10.57005/ab.2026.2.9
Adaptive reinforcement learning based projected gradient descent attack
Zhu, Z. D. J. D., Zhang, Y., Ye, A., Wang, X. Y., Wang, C., & Zhu, T. (2026) · The Journal of Supercomputing · doi:10.1007/s11227-026-08687-z · Cited by 1
RolePlayAI: A Multi-agent Debate Framework for Creative Design
Zhang, W., Xiao, Y., & Yu, H. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-30846-7_14
TaxAgent-MA: A Multi-Agent framework for Auditable and Explainable Tax Recommendation Generation
Gao, Y., Cao, J., xie, Z., Wang, L., & Yang, W. Z. (2026) · International Conference on Computer Supported Cooperative Work in Design (CSCWD) · doi:10.1109/cscwd68734.2026.11582797
“Representative works such as AutoGen, CAMEL, and LangGraph [14], [15] achieve high interpretability and controllability through role-based interaction and iterative dialogue.”
KBC-Coop: Knowledge-Grounded Role Cooperation for Single-Model LLMs in Collaborative Tasks
Yang, H. Q., Wu, L., Lin, S., Wang, Z., & Keung, J. (2026) · International Conference on Computer Supported Cooperative Work in Design (CSCWD) · doi:10.1109/cscwd68734.2026.11582521
“…reasoning [6], MetaGPT encodes standard operating procedures into a pipeline of role-structured agents for software engineering [7], CAMEL uses role-playing agents to study cooperation and social behavior [8], and Generative Agents simulate long-term social interaction in shared environments [9].”
LLMaaD: LLM-Based Multi-Agent Framework for Interpretable Log Anomaly Detection
Yin, H., Chen, H., Zhang, R., & Zhang, K. (2026) · International Conference on Computer Supported Cooperative Work in Design (CSCWD) · doi:10.1109/cscwd68734.2026.11581712
“Similarly, while generic multi-agent frameworks like CAMEL [19] enable collaborative problem-solving, they lack domain-specific mechanisms for heterogeneous AIOps data.”
Lee, N., & Kwon, O. (2026) · Applied Sciences · doi:10.3390/app16136715 · Cited by 1
“LLMs can be designed as artificial agents capable of generating alternatives, evaluating options, summarizing opinions, and processing feedback through natural language interaction [6].”
LGAgent: A Structured and Rule-Guided Multi-Agent Framework for Interpretable Legal ReasoningPreprint
Feng, X., Shu, F., Feng, M., Feng, Y., & Liu, H. F. (2026) · Research Square · doi:10.21203/rs.3.rs-9571992/v1
Li, Y., Sun, Q., Xue, J., Javadov, A., Milling, M., Zhang, W. H., Kainz, B., & Schuller, B. W. (2026) · Frontiers in Digital Health · doi:10.3389/fdgth.2026.1803433
“For example, CAMEL [23] initialises each agent using role-playing prompts while sharing model parameters.”
ALBA: A Multi-agent System Based on Large Language Models for Business Process Automation
Vrljić, B., Botički, I., & Poje, K. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-20897-2_40
Rani, D. B. A. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-35586-7_2 · Cited by 1
CogMed: A Multi-Agent Legal Mediation Framework Fusing Cognitive Strategies and Dynamic Beliefs
Chen, J., Ma, Y., & Gao, S. (2026) · Information · doi:10.3390/info17070671
Divi, V. R., Gullapalli, S., & Brahmandam, B. A. (2026) · International Conference on Electronics Representation and Algorithm (ICERA) · doi:10.1109/icera72709.2026.11666599 · Cited by 1
“AutoGen [4], MetaGPT [5], CAMEL [6], ChatDev [7], and AgentVerse [12] popularized role-based collaboration and orchestrator-worker patterns.”
Jinuk, L. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-2862-1_29
Zhu, T., Yang, L., Li, L., Tian, Y. W., & Xu, H. (2026) · Sustainability · doi:10.3390/su18147128
Chen, W., Sun, K., Wen, J., Zhu, Y., & Fang, J. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3444-8_19
Xue, R., Sun, R., Altarrazi, S. M., Jha, D. N., Li, Y., Wealls, P., Szydło, T., Dustdar, S., & Ranjan, R. (2026) · IEEE Internet Computing · doi:10.1109/mic.2026.3662039
Zhang, X., Nie, X., Liu, Y., & Wang, Z. (2026) · Mathematics · doi:10.3390/math14142562
StoryCheck: A Scalable Framework for Automated Evaluation of Creative Writing using AI Agents
Shittu, L., Akintayo, M., Arabambi, A., Okeleye, F., Owah, O. L., Akinfenwa, O., Olajide-Shokunbi, A., Jaiyesimi, B. S., Fagbohun, O., & Adetunji, R. A. (2026) · Cognitive Models and Artificial Intelligence Conference (AICCONF) · doi:10.1109/aicconf69182.2026.11600738
“Li et al. [17] introduced CAMEL, a framework for studying cooperative agent behaviour through role-playing, which has implications for understanding how multiple perspectives can be synthesised into coherent judgments.”
A Multi-agent System Integrating LLM for Intelligent Athlete Assistance
Costa, A., Oliveira, P., Magalhães, R., Nováis, P., & Durães, D. A. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-32029-2_27 · Cited by 1
Huang, T., Zhang, J., Yan, P., Yang, R., & Li, X. (2026) · International Conference on Artificial Intelligence and Materials (ICAIM) · doi:10.1109/icaim69488.2026.11601646
“Rule-Based and Designated Orders enforce predetermined or speaker-chosen sequences, which can restrict adaptability and marginalise less active agents [17].”
Closed-Loop Recovery Mechanism for Multi-agent Process Planning Under Manufacturing Constraints
Yin, B., & Sun, N. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3403-5_13
BICA-Inspired Multiagent Model of Self-Awareness Based on LLM
Samsonovich, A. V., & Khabarov, D. L. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-33195-3_19
Dynamic M2CL: Process-Level Convergence Control for LLM-Based Multi-Agent Deliberation
Wei, Q., & Liu, Y. (2026) · International Conference on Complex Systems and Automation Control (CSAC) · doi:10.1109/csac70128.2026.11607501
“LLM-based multi-agent reasoning has been studied through debate, self-consistency, role-based collaboration, and agent societies, including Multiagent Debate [5], Self-Consistency [7], and frameworks such as AutoGen, CAMEL, ChatDev, and Generative Agents [6], [8], [9].”
MACE: A Physics-Informed Multi-modal Large Language Model for Strategic UAV Swarm Confrontation
Yang, Y., Shi, H. H., & Zhu, J. (2026) · International Conference on Robotics and Control Engineering (RobCE) · doi:10.1145/3820709.3820744
“However, these systems primarily operate on structured textual abstractions [16], which inevitably filter out high-dimensional situational features such as terrain gradients and acoustic threat orientations.”
A multi-agent framework powered by large language models for automatic heuristic design
Yang, J., Chen, X., Qu, R., Qiu, Z., & Bai, R. (2026) · Swarm and Evolutionary Computation · doi:10.1016/j.swevo.2026.102487
Virtual Roundtable: Multi-Agent Personas Simulating the Dynamics of Human Brainstorming
Dorn, T., Khan, S. A., & Mumford, J. (2026) · International Conference on Human-AI Interaction and Experience Design (HAXD) · doi:10.1109/haxd70072.2026.11620818
“CAMEL [12] introduced cooperative role-playing between two domain-specialized agents for task completion, and Wang et al. [21] extended this to multi-persona self-collaboration.”
OmegaClaw: A Continually Operating Agentic Architecture Under Bounded Resources
Botnick, M., Hammer, P., Isaev, P., Goertzel, B., & Crawford, K. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-33010-9_22
MAGen: Multi-agent smart contract generation with automated testing and verification
Liu, L., Ke, W., Chi, H., & Yan, J. (2026) · Empirical Software Engineering · doi:10.1007/s10664-026-10931-2
“We utilize the inception prompting technique (Li et al. 2023 ), by which the system enables agents to effectively assume and fulfill their designated roles.”
Web Log Anomaly Detection via Multi-Agent RAG
Hao, Q., & Wang, X. (2026) · International Conference on Big Data & Artificial Intelligence & Software Engineering (ICBASE) · doi:10.1109/icbase70763.2026.11619482
“To break the cognitive limitations of standalone architectures and address the correlation loss from retrieval isolation, this paper designs a role-playing-based red-blue multi-agent adversarial mechanism [15].”
Agentic Metaverse Services: A New As-a-Service Paradigm
Xu, X., Sheng, Q. Z., Wang, Z., Benatallah, B., Wang, X., & Han, R. (2026) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws72778.2026.00162
“The platforms such as AutoGen [28] and CAMEL [29] provided general-purpose multi-agent platforms for dialogue orchestration and role-based cooperation, while Sapkota et al. [30] clarified the distinction between monolithic AI agents and multi-agent agentic AI systems.”
Yan, P. S., Li, H., Chen, N., Li, L., & Chen, P. (2026) · Theory and Society · doi:10.1007/s11186-026-09735-6 · Cited by 1
“Example applications include multi-agent role-play (for example, CAMEL), where two or more role-conditioned agents socialise, cooperate or compete through open-ended dialogue using prompting (Li et al., 2023).”
A New Paradigm: Agentic AI for Scientific Discovery
Taktakidze, A. (2026) · Longevity Horizon · doi:10.65649/ukh4n33x · Cited by 1
“…2023); language models augmented with the ability to call external tools — databases, APIs, code interpreters (Schick et al., 2023; Qin et al., 2023); and multi-agent orchestration frameworks that coordinate specialised components into collaborative workflows (Li et al., 2023; Park et al., 2023).”
Platform Sociality: Mediality as a Condition of Agency
Tikhonova, S. V., & Kornievsky, A. A. (2026) · Galactica Media Journal of Media Studies · doi:10.46539/gmd.v8i3.746
Luo, J., Liu, M., Huang, Z., Ke, Z., & Wang, Q. (2026) · Electronics · doi:10.3390/electronics15153452
EzBlender: Efficient 3D Editing with Plan-and-ReAct Agent
Wang, H., Zhu, W., Tang, S., Wang, Z., Dong, X., Li, X., Chen, X., Bastola, A., Huang, X., Wang, Y., et al. (2026) · IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw68408.2026.00146 · Cited by 1
“Systems such as Manus[34], ChatGPT Agents[38], AutoGen[48], and CAMEL[25] go well beyond single-task execution, demonstrating that modern agents are evolving into flexible ecosystems rather than fixed patterns, capable of coordinating multiple roles and handling intricate workflows.”
Agentic Foresight: Potenziale autonomer KI-Agenten für die strategische Vorausschau in Unternehmen
Haugk, S., & Leyh, C. (2026) · HMD Praxis der Wirtschaftsinformatik · doi:10.1365/s40702-026-01318-4
Rai, B., & Popović, M. J. (2026) · Applied Decision Analytics · doi:10.66972/ada202734
Lee, J., & Li, W.-S. (2026) · IEEE Access · doi:10.1109/access.2026.3721147 · Cited by 1
“…schemes such as Chain-of-Agents, further show that distributing context across specialized or sequential agents is a live design pattern, even though they do not make our specific claim about temporary scoped projection with provenance-preserving rejoin inside one trace graph [28], [29], [30].”
StepFinder: A Temporal Semantic Framework for Failure Attribution in Multi-Agent Systems
Zhu, T., Wu, Y., Jin, W., Li, Y., & Huang, G. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770855.3817991 · Cited by 6
“Consequently, MAS consistently outperform single-agent systems in complex scenarios such as program synthesis and scientific discovery [10, 19, 28, 31].”
Iklassov, Z., Madmoun, H., Duhot, J.-J., & Lahlou, S. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770855.3816459
Parallelizing LLM Agent Execution with Contrastive Task Allocation
Peng, Y., Xu, Y., Wang, S., Liao, X., & Li, Q. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770855.3817872 · Cited by 2
“Whilesingle-agentworkflowsevolve,Multi-Agent Systems(MAS) like AutoGen [34], MetaGPT [15], and CAMEL [20] have popu-larized collaborative problem-solving through role-playing.”
Handoff Hallucinations: Taxonomy, Benchmark, and Mitigation for Multi-agent Pipeline Failures
Jafarova, H., & Alakbarli, F. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-31319-5_22
Agentic Ecosystem for Question Answering using Multi-Persona Agents
Vrisheeka, M., Mandlik, A. J., Singh, P. K., Kavisankar, L., & Ray, B. (2026) · International Conference on Data Science and Business Systems (ICDSBS) · doi:10.1109/icdsbs69077.2026.11634648
“The research work [2] offers a framework based on roles and agent communication allowing structured conversations; however, it lacks any supervision tools for synthesizing responses and evaluating results. for tokenization, generation, and interpretation, which is the NLU layer.”
POSTER: ActShare: Coordinating Reusable Action Requests in Multi-Agent Workflows
Kang, R., Zhu, M., & Li, T. (2026) · ACM SIGCOMM Conference (SIGCOMM) · doi:10.1145/3789240.3830289
“Systems such as AutoGen, CAMEL, and MetaGPT use role-based collaboration to improve task decomposition and execution flexibility [3, 5, 10], while workflow optimization often represents agentic workflows as structured graphs or code-defined search spaces [11, 12].”
Qi, Z., Saito, H., Koseki, R., Suzuki, M., Kubota, S., Naito, T., Inaba, M., Kameue, C., Matsukura, H., & Sakamoto, M. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-35387-0_33
A survey of reward hacking in agentic large language model systems
Morampudi, A., Irrink, U., Grandhi, R., Pagadala, V., & Maddula, M. (2026) · Discover Artificial Intelligence · doi:10.1007/s44163-026-01980-z · Cited by 1
“Key architectural developments include Toolformer (self-supervised tool use) [34], ReAct (reasoning-action interleaving) [35], CAMEL (multi-agent coordination) [36], generative agents (persistent memory and planning) [37], and CoALA (cognitive architecture for language agents) [38].”
RoCoMAP: A Role-Based Collaborative Multi-Agent Framework for Patent Supply–Demand Matching
He, X., An, J., Zhang, Z., & Ni, X. (2026) · Scientometrics · doi:10.1007/s11192-026-05777-w
“Frameworks like CAMEL (Li et al., 2023), MARE (Jin et al., 2024), and FinCon (Yu et al., 2024) further demonstrate this potential by applying role-playing in software development, cooperative structures in requirements engineering, and manager–analyst hierarchies in financial decision-making.”
Explicit Cognitive Allocation: A Principle for Governed and Reviewable LLM-Assisted Inquiry
Manzanilla-Granados, H. M., Cazales, Z. N., Pescador‐Rojas, M., & Ramírez-Romero, T. A. (2026) · IEEE Access · doi:10.1109/access.2026.3725913
“Multi-agent frameworks such as CAMEL, AutoGen, and MetaGPT organize task execution through role-playing, conversable agents, human participation, tool use, and standardized operating procedures [24], [25], [26].”
Congestion-Aware Scheduling for Heterogeneous LLM-Agent Teams
Xie, W., & Li, Y. (2026) · IEEE Annual International Computer Software and Applications Conference (COMPSAC) · doi:10.1109/compsac69091.2026.00432
“CAMEL studies role-playing communication among agents [14], while SPP turns one LLM into a multi-persona collaborator [15].”
Okeleye, F., Oluwaseun, B., Fagbohun, O., Sikiru, O., Akindahunsi, T., Jaiyesimi, B. S., Adetunji, R. A., Arabambi, A., & Akintola, A. S. (2026) · IEEE International Conference on Big Data and Artificial Intelligence (BDAI) · doi:10.1109/bdai70753.2026.11655102
“The rise of LLM agents [15], [16] and frameworks such as AutoGen [17], MetaGPT [18 CAMEL [19], and AgentVerse [20] demonstrate that role specialization and structured dialogue improve collaborative reasoning, but applications to document-intensive financial due diligence remain limited.”
Communication, Embodiment, and Coordination in an El Farol Bar-Inspired LLM Society
Takata, R., Masumori, A., & Ikegami, T. (2026) · Artificial Life · doi:10.1162/artl.a.459
Zhao, Z., Guo, Y., Liu, X., Ma, Q., Yu, H., & Xue, X. (2026) · Expert Systems · doi:10.1111/exsy.70392
“On each environment we run four open-source agent frameworks under their public prompt templates: ReAct (Yao et al. 2023) with a single thought–action step, AutoGen (Wu et al. 2024) with a planner–critic pair, Generative Agents (Park et al. 2023) with reflection before action, and CAMEL (Li et al. 2023) with a user–as…”
InfluentialSemantic Raft: A Fault-Tolerant Multi-Agent Inference Framework for Reliable LLM Services
Su, Y.-D., Wang, Y., Dong, Y., & Xu, Z. (2026) · International Workshop on Quality of Service · doi:10.1109/iwqos70441.2026.11661187
“Frameworks like CAMEL [23] and MetaGPT [24] utilize role-playing agents for task decomposition, while Collaborative Beam Search [25] and Multi-Agent Debate (MAD) [13] optimize answers using peer feedback.”
Formal Trajectory Analysis for Testing Agentic AI in Stateful Environments
Twabi, A., Ding, Y., & Kondo, T. (2026) · IEEE International Conference on Artificial Intelligence Testing (AITest) · doi:10.1109/aitest70988.2026.00025
“A growing ecosystem of multi-agent frameworks supports diverse coordination patterns, including conversation-based orchestration [31], role-based pipelines [32]–[34], and graph-based workflows [35].”
FTDI: A budget-aware self-healing framework for LLM multi-agent code generation
Men, S., Lu, Z., Luo, H., & Zheng, Y. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.134164
DelAct: A Replayable Boundary Runtime for Auditable and Governed LLM Agent Workflows
Zhang, Y., Liao, H., Guo, D., & Tang, G. (2026) · International Workshop on Quality of Service · doi:10.1109/iwqos70441.2026.11661202
Condori-Bustincio, K. F., Paredes-Cayllahua, E. L., Bustincio, R. W. C., Junior, W. V. L., & Bittencourt, L. F. (2026) · International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT) · doi:10.1109/dcoss-iot69657.2026.00191
“Recent LLM-based simulations extend this paradigm by using language models as generative agents capable of producing socially plausible actions, plans, and dialogues [38]–[43].”
WaSMa: Workflow-Aware Scheduling for Multi-Agent LLM Systems
Rao, U., Sadiq, A. Z., Shen, H., Yang, Y., Hodges, E.-A. R., & Muntean, M. (2026) · International Conference on Computer Communications and Networks (ICCCN) · doi:10.1109/icccn69946.2026.11662938
“Assisted by the availability of popular frameworks such as AutoGen [18], MetaGPT [5], CAMEL [9], AgentVerse [3], ChatDev [15], and DSPy [7], these applications increasingly employ multiple AI agents or structured LM programs that interact through natural language prompts to complete complex tasks.”
Zhang, Y., Zhou, P., Zhang, L., & Wang, R. (2026) · Aerospace Systems · doi:10.1007/s42401-026-00487-w
Zhou, X., & Jiang, Z. (2026) · Applied System Innovation · doi:10.3390/asi9090179
DiagAgent: A LLM-Based Multi-Agent Framework for Highly Accurate Product Fault Diagnosis
Zhang, H., Xu, K., & Tang, Y. (2025) · International Symposium on Computational Intelligence and Design · doi:10.1109/iscid68789.2025.00044
“Prompt engineering[5] is one of the most commonly used meth-ods, where researchers design customized role-specific prompts[6], task templates, and behavioral constraints for each agent, enabling different agents to have clear divisions of labor and interaction norms.”
ClaimPilot: An Open Multi-Agent Pipeline for Veterinary Insurance Claims Adjudication
Hoang, M. L., & Shunmugam, N. (2026) · Future Internet · doi:10.3390/fi18090465
Large language models in contexts of uncertainty: the “digital habitus” framework
Olivieri, M. G., & Iavarone, L. (2026) · Quality & Quantity · doi:10.1007/s11135-026-03049-6
Role-Aware Agentic Retrieval-Augmented Generation for Urban Rail Transit Fault Management
Zhang, W., Fang, H., Bu, J., Zhong, Q., & Yu, Y. (2026) · AI · doi:10.3390/ai7090339
“More recently, multi-agent frameworks, such as CAMEL and MetaGPT, have demonstrated that decomposing complex tasks into coordinated role-specific agents can improve scalability and collaborative problem solving [22,23].”
Discharge-agents: A few-shot multi-agent collaborative framework for discharge summary generation
Che, Z.-H., Chen, Z., & Xu, Y. (2026) · Pattern Recognition Letters · doi:10.1016/j.patrec.2026.09.003
Kim, S., Noh, K., Choi, B. K., & Hahn, H. (2026) · Journal of Society of Korea Industrial and Systems Engineering · doi:10.11627/jksie.2026.49.2.095
Engert, G. E., Bischoff, A.-K., & Weigold, M. (2026) · Procedia CIRP · doi:10.1016/j.procir.2026.03.275
Contagion on the Trading Floor: How Adversarial Signals Spread in Multi-Agent Trading Systems
Rong, S. Q., Dong, J., Thai, N. D., Wen, Y., Tan, C., & Ong, Y.-S. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37673-2_12 · Cited by 1
“The rapid rise of large language models (LLMs) has led to a corresponding surge in LLM-based agents and multi-agent systems [28,13,10,21,9].”
RoleCraft: Aligning Persona Stance and Style for Robust Role-Playing Agents
Gao, Y., Dang, C., He, Y., Hu, J., Luo, Y., & Yang, J. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37667-1_9
Early Find Early Stop: Enhancing Multi-Agent System Efficiency Via Hierarchical Real-Time Validation
Ma, Z., Shu, Y., Zhang, L., Chen, D., Xia, F., Zhang, W., Zuo, D., & Wen, D. (2026) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws72778.2026.00119
“By decomposing complex problems into subtasks [2], agents with specialized roles can collaborate to achieve capabilities far beyond those of a single model [3], [4].”
Kumar, N. A., & Kataria, V. (2026) · Journal of Hospitality and Tourism Insights · doi:10.1108/jhti-12-2025-1458
Large-language-model-empowered Agents for Aerospace Tool Kit Applications
Zhang, J., ZHANG, J., Hua, W. A. N. G., & Yazhong, L. U. O. (2026) · Manned Spaceflight · doi:10.3724/zrht.1674-5825.2026053
Shu, Y., Yan, W., Ma, Z., Yang, J., Jiang, Y., Li, M., Luo, T., & Zuo, D. (2026) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws72778.2026.00157
“In many existing approaches, all agents participate in discussion for every task [2].”
Li, Z., Wen, J., Liao, J., Jin, X., Miao, X., Pan, G., Zhong, W., Xie, S., & Kim, D. I. (2026) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2026.3731984
“…through standardized protocols such as the Agent-to-Agent (A2A) protocol [12] to perform complex tasks including capability discovery [13], [14], task negotiation [15], [16], and cross-domain collaboration [17], [18], without relying on a pre-established central authority for trust management.”
Multi-Agent LLM Collaborative Reasoning and Task Planning for Complex Task Solving
Zhang, W., Liao, T., & Feng, Y. (2026) · International Conference on Computer Information Science and Application Technology (CISAT) · doi:10.1109/cisat71030.2026.11682610
“CAMEL explored role-playing communicative agents for autonomous cooperation, highlighting the value of role assignment and inception prompting [13].”
LLM‐Powered AI Agents in Materials Science: From Autonomous Design to Autonomous Experimentation
Jiang, T., Zhang, Z., Nie, M., Zhu, D., Meng, E., Wu, H., & Mao, X. (2026) · Advanced Functional Materials · doi:10.1002/adfm.78275 · Cited by 1
A Flexible Multi-Agent LLM-Human Framework for Fast Human Validated Tool Building
Daull, X., Bellot, P., Bruno, E., Martin, V., & Murisasco, É. (2025) · IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT) · doi:10.1109/wi-iat67162.2025.00058
“Those approaches extended CAMEL’s role-playing agents [13] to cooperate through inception-prompted dialogue to achieve complex tasks, sequential skills (tools) acquisition [4], mathematical discovery [5], chain-of-thought reasoning [8], and multi-agent coordination [11].”
Li, D., & Liu, Y. (2026) · International Conference on Cloud Computing and Big Data Analytics (ICCCBDA) · doi:10.1109/icccbda69573.2026.11676497
“General frameworks such as CAMEL [14] and AutoGen [15] have shown the potential of agents in instruction following.”
Więckowski, J., Paradowski, B., Kizielewicz, B., & Bouraima, M. B. (2026) · Applied Soft Computing · doi:10.1016/j.asoc.2026.116417 · Cited by 1
AGE: Agentic Gaussian Editing in 3D Scenarios
Qin, H., Lin, T., Wei, M., Sun, Y., Lu, M., Kong, M., & Zhu, Q. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37553-7_13
Xiang, Y. (2026) · International Conference on Large Models and Intelligent Agent Technologies (LMIAT) · doi:10.1109/lmiat70705.2026.11682229 · Cited by 2
“Decentralized and market-based allocation revives the classical contract-net protocol [25], letting agents bid for tasks; CAMEL’s role-playing dialogue [9] and AgentVerse’s dynamic recruitment [10] sit here, scaling better but weakening any global invariant.”
Metadata
- Authors
- Li, G., Hammoud, H., Itani, H., Khizbullin, D., & Ghanem, B.
- Venue
- Advances in Neural Information Processing Systems 36
- Published
- 2023
- Pages
- 51991-52008
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W7133220405
- Cited by
- 314
- References
- 0
- Citations by year
- 2026: 197 · 2025: 77 · 2024: 32 · 2023: 8
- FWCI
- 78.60
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- cc-by
- Locations
- neural information processing systems (publishedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- King Abdullah University of Science and Technology
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
- Physical Sciences › Computer Science › Artificial Intelligence › Multi-Agent Systems and Negotiation
- Keywords
- large language models, multi-agent systems, agent communication, emergent behavior, collective intelligence
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.