Abstract
Conversational systems are becoming a primary interface for services and enterprise automation, and rapid market growth is pushing deployments into safety- and cost-sensitive settings. Reliability remains a bottleneck when interactions span multiple domains: an orchestrator must choose the next specialist, maintain shared dialogue state, and recover from mistakes before they cascade across handoffs. Despite rising interest in swarm-like multi-agent designs, orchestration is rarely evaluated with coordination-centric metrics, making it hard to compare routing policies beyond surface fluency. We present an evaluation-first pipeline for multi-domain task-oriented dialogue on MultiWOZ 2.2 that decouples routing from generation and exposes measurable failure modes. A DeBERTa-based router selects domain specialists, while a FLAN-T5 generator produces structured actions and belief-state updates under a shared memory interface. The protocol tracks delegation correctness, slot-progress coverage, switching and bouncing instability, loop behavior, and recovery after misroutes, and it links early-turn errors to downstream collapse using cascading-error attribution. We further introduce stress tests that simulate reformulation, long-horizon corrections, and tool-latency delays to probe robustness beyond static annotations. Across routing variants, confidence-aware gating yields the strongest stability improvement, achieving routing accuracy of 0.77 while substantially reducing handoff churn, with switching 0.11 and bounce 0.01, relative to a learned baseline with 0.65 accuracy, switching 0.44, and bounce 0.09. At the same time, confidence gating can trade progress for precision when it suppresses belief updates, highlighting an accuracy-progress tension that is important for deployment tuning. Diagnostic summaries identify misrouting and empty-state updates as dominant contributors, while looping is comparatively rare. Finally, applying the same evaluation to SGD shows that coordination challenges persist under schema shift. Overall, the proposed metrics and implementation blueprint provide a reproducible basis for diagnosing coordination failures and selecting orchestration policies for deployment.
Data availability
The MultiWOZ 2.2 and Schema-Guided Dialogue (SGD) datasets, along with the complete source code used in this study, are publicly available at the following links: \(\bullet\) Dataset: MultiWOZ 2.2 Dataset \(\bullet\) Dataset: Schema-Guided Dialogue (SGD) Dataset \(\bullet\) Source Code: GitHub Repository
References
Chong, T., Yu, T., Keeling, D. I. & de Ruyter, K. AI-chatbots on the services frontline addressing the challenges and opportunities of agency. J. Retail. Consum. Serv. 63, 102735. https://doi.org/10.1016/j.jretconser.2021.102735 (2021).
Kietzmann, J. & Park, A. Written by ChatGPT: AI, large language models, conversational chatbots, and their place in society and business. Bus. Horiz. 67, 453–459. https://doi.org/10.1016/j.bushor.2024.06.002 (2024).
Estevez, M., Ballestar, M. T. & Sainz, J. Market research and knowledge using generative AI: The power of large language models. Journal of Innovation & Knowledge 10, 100796. https://doi.org/10.1016/j.jik.2025.100796 (2025).
Ferraro, C., Demsar, V., Sands, S., Restrepo, M. & Campbell, C. The paradoxes of generative AI-enabled customer service: A guide for managers. Bus. Horiz. 67, 549–559. https://doi.org/10.1016/j.bushor.2024.04.013 (2024).
Hermann, E. & Puntoni, S. Artificial intelligence and consumer behavior: From predictive to generative AI. J. Bus. Res. 180, 114720. https://doi.org/10.1016/j.jbusres.2024.114720 (2024).
Kwan, W.-C., Wang, H.-R., Wang, H.-M. & Wong, K.-F. A survey on recent advances and challenges in reinforcement learning methods for task-oriented dialogue policy learning. Mach. Intell. Res. 20, 318–334. https://doi.org/10.1007/s11633-022-1347-y (2023).
Maroengsit, W., Piyakulpinyo, T., Phonyiam, K. & Theeramunkong, T. A survey on evaluation methods for chatbots. In Proceedings of the 7th International Conference on Information Technology (InCIT 2019), 1–6, https://doi.org/10.1145/3323771.3323824 (ACM, New York, NY, USA, 2019).
Sapkota, R., Roumeliotis, K. I. & Karkee, M. AI agents vs. agentic AI: A conceptual taxonomy, applications and challenges. Inf. Fusion 126, 103599. https://doi.org/10.1016/j.inffus.2025.103599 (2026).
Althaf, A. M., Mohammed, M. A., Milanova, M., Talburt, J. & Cakmak, M. C. Multi-agent RAG framework for entity resolution: Advancing beyond single-LLM approaches with specialized agent coordination. Computers 14, 525. https://doi.org/10.3390/computers14120525 (2025).
Vázquez, A., López Zorrilla, A., Olaso, J. M. & Torres, M. I. Dialogue management and language generation for a robust conversational virtual coach: Validation and user study. Sensors 23, 1423. https://doi.org/10.3390/s23031423 (2023).
Liesenfeld, A. & Dingemanse, M. Interactive probes: Towards action-level evaluation for dialogue systems. Discourse Commun. 18, 954–964. https://doi.org/10.1177/17504813241267071 (2024).
Ohashi, A. & Higashinaka, R. Optimizing pipeline task-oriented dialogue systems using post-processing networks. Comput. Speech Lang. 90, 101742. https://doi.org/10.1016/j.csl.2024.101742 (2025).
Deriu, J. et al. Survey on evaluation methods for dialogue systems. Artificial Intelligence Review 54, 755–810. https://doi.org/10.1007/s10462-020-09866-x (2021).
Yi, Z. et al. A survey on recent advances in llm-based multi-turn dialogue systems. ACM Computing Surveys 58, 1–38. https://doi.org/10.1145/3771090 (2025).
Razumovskaia, E. et al. Crossing the conversational chasm: A primer on natural language processing for multilingual task-oriented dialogue systems. Journal of Artificial Intelligence Research 74, 1351–1402. https://doi.org/10.1613/JAIR.1.13083 (2022).
Ni, J., Young, T., Pandelea, V., Xue, F. & Cambria, E. Recent advances in deep learning based dialogue systems: A systematic survey. Artif. Intell. Rev. 56, 3055–3155. https://doi.org/10.1007/s10462-022-10248-8 (2023).
Lee, H., Jo, S., Kim, H., Jung, S. & Kim, T.-Y. SUMBT+LaRL: Effective multi-domain end-to-end neural task-oriented dialog system. IEEE Access 9, 116133–116146. https://doi.org/10.1109/ACCESS.2021.3105461 (2021).
Heck, M. et al. Robust dialogue state tracking with weak supervision and sparse data. Trans. Assoc. Comput. Linguist. 10, 1175–1192. https://doi.org/10.1162/tacl_a_00513 (2022).
Liao, L., Long, L. H., Ma, Y. & Chua, T.-S. Dialogue state tracking with incremental reasoning. Trans. Assoc. Comput. Linguist. 9, 557–569. https://doi.org/10.1162/tac0l_a_00384 (2021).
Li, J., Song, S. & Yan, S. Advanced dialog state tracking with noetic graphs for complex human-machine interactions. Pattern Recogn. 168, 111842. https://doi.org/10.1016/j.patcog.2025.111842 (2025).
Khan, M. A. et al. A multi-attention approach using bert and stacked bidirectional lstm for improved dialogue state tracking. Appl. Sci. 13, 1775. https://doi.org/10.3390/app13031775 (2023).
Yu, H. & Ko, Y. Enriching the dialogue state tracking model with a asyntactic discourse graph. Pattern Recognit. Lett. 169, 81–86. https://doi.org/10.1016/j.patrec.2023.03.024 (2023).
Lu, H. et al. Prompt-based end-to-end cross-domain dialogue state tracking. Electronics 13, 3587. https://doi.org/10.3390/electronics13183587 (2024).
Tsinganos, N., Fouliras, P. & Mavridis, I. Leveraging dialogue state tracking for zero-shot chat-based social engineering attack recognition. Appl. Sci. 13, 5110. https://doi.org/10.3390/app13085110 (2023).
Hong, T., Cho, J., Yu, H., Ko, Y. & Seo, J. Knowledge-grounded dialogue modelling with dialogue-state tracking, domain tracking, and entity extraction. Comput. Speech Lang. 78, 101460. https://doi.org/10.1016/j.csl.2022.101460 (2023).
Jia, X., Zhang, R. & Peng, M. Multi-domain gate and interactive dual attention for multi-domain dialogue state tracking. Knowl. Based Syst. 286, 111383. https://doi.org/10.1016/j.knosys.2024.111383 (2024).
Xi, Z. et al. The rise and potential of large language model based agents: A survey. Sci. China Inf. Sci. 68, 121101. https://doi.org/10.1007/s11432-024-4222-0 (2025).
Wang, L. et al. A survey on large language model based autonomous agents. Front. Comput. Sci. 18, 186345. https://doi.org/10.1007/s11704-024-40231-1 (2024).
Qu, C. et al. Tool learning with large language models: A survey. Front. Comput. Sci. 19, 198343. https://doi.org/10.1007/s11704-024-40678-2 (2025).
Li, X., Wang, S., Zeng, S., Wu, Y. & Yang, Y. A survey on LLM-based multi-agent systems: Workflow, infrastructure, and challenges. Vicinagearth 1, 9. https://doi.org/10.1007/s44336-024-00009-2 (2024).
Gao, C. et al. Large language models empowered agent-based modeling and simulation: A survey and perspectives. Humanit. Soc. Sci. Commun. 11, 1259. https://doi.org/10.1057/s41599-024-03611-3 (2024).
Wang, Y. et al. Large model based agents: State-of-the-art, cooperation paradigms, security and privacy, and future trends. IEEE Commun. Surv. Tutor. https://doi.org/10.1109/COMST.2025.3576176 (2025). Accepted/In press.
Liu, Y. et al. Datasets for large language models: A comprehensive survey. Artif. Intell. Rev. 58, 403. https://doi.org/10.1007/s10462-025-11403-7 (2025).
Lee, P., Son, M. & Jia, Z. Ai-powered automatic item generation for psychological tests: A conceptual framework for an llm-based multi-agent aig system. J. Bus. Psychol. https://doi.org/10.1007/s10869-025-10067-y (2025).
Song, A. & Azman, A. Enhancing LLM-driven multi-agent code generation through cross verification and joint optimization. Symmetry (Basel) 17, 1660. https://doi.org/10.3390/sym17101660 (2025).
Perera, R., Basnayake, A. & Wickramasinghe, M. Auto-scaling LLM-based multi-agent systems through dynamic integration of agents. Front. Artif. Intell. 8, 1638227. https://doi.org/10.3389/frai.2025.1638227 (2025).
Piccialli, F. et al. Agentai: A comprehensive survey on autonomous agents in distributed AI for Industry 4.0. Expert Syst. Appl. 291, 128404. https://doi.org/10.1016/j.eswa.2025.128404 (2025).
Abou Ali, M., Dornaika, F. & Charafeddine, J. Agentic AI: A comprehensive survey of architectures, applications, and future directions. Artif. Intell. Rev. https://doi.org/10.1007/s10462-025-11422-4 (2026).
Xie, J., Chen, Z., Zhang, R. & Li, G. Large multimodal agents: A survey. Vis. Intell. https://doi.org/10.1007/s44267-025-00093-y (2025).
Xia, C. S., Deng, Y., Dunn, S. & Zhang, L. Demystifying LLM-based software engineering agents. Proc. ACM Softw. Eng. 2, 801–824. https://doi.org/10.1145/3715754 (2025).
Kondylidis, N., Tiddi, I. & ten Teije, A. A framework for establishing shared, task-oriented understanding in hybrid open multi-agent systems. Front. Artif. Intell. 8, 1440582. https://doi.org/10.3389/frai.2025.1440582 (2025).
Legashev, L., Shukhman, A., Badikov, V. & Kurynov, V. Using large language models for goal-oriented dialogue systems. Applied Sciences 15, 4687. https://doi.org/10.3390/app15094687 (2025).
Sun, J., Kou, J., Shi, W. & Hou, W. A multi-agent collaborative algorithm for task-oriented dialogue systems. Int. J. Mach. Learn. Cybern. 16, 2009–2022. https://doi.org/10.1007/s13042-024-02374-2 (2025).
squiduu. Multiwoz 2.2. Kaggle dataset (2022). Updated Jan 30, 2022. Accessed: 2026–02-08.
google-research-datasets. dstc8-schema-guided-dialogue. GitHub repository (2019). Schema-Guided Dialogue (SGD) and SGD-X datasets; CC BY-SA 4.0. Accessed: 2026–02-08.
Rastogi, A., Zang, X., Sunkara, S., Gupta, R. & Khaitan, P. Towards scalable multi-domain conversational agents: The schema-guided dialogue dataset. Proc. AAAI Conf. Artif. Intell. 34, 8689–8696 (2020).
Wu, C.-S. et al. Transferable multi-domain state generator for task-oriented dialogue systems. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 808–819, https://doi.org/10.18653/v1/P19-1078 (Association for Computational Linguistics, 2019).
Zang, X. et al. Multiwoz 2.2: A dialogue dataset with additional annotation corrections and state tracking baselines. In Proceedings of the 2nd Workshop on Natural Language Processing for Conversational AI, 1442, https://doi.org/10.18653/v1/2020.nlp4convai-1.13 (Association for Computational Linguistics, 2020).
Zhang, J. et al. Find or classify? dual strategy for slot-value predictions on multi-domain dialog state tracking. In Proceedings of the Ninth Joint Conference on Lexical and Computational Semantics, 154–164 (Association for Computational Linguistics, 2020).
Hosseini-Asl, E., McCann, B., Wu, C.-S., Yavuz, S. & Socher, R. A simple language model for task-oriented dialogue. arXiv:2005.00796. (2020).
Tian, X. et al. Amendable generation for dialogue state tracking. In Proceedings of the 3rd Workshop on Natural Language Processing for Conversational AI, 62–70, https://doi.org/10.18653/v1/2021.nlp4convai-1.8 (Association for Computational Linguistics, 2021).
Feng, Y., Wang, Y. & Li, H. A sequence-to-sequence approach to dialogue state tracking. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), 1691–1701, https://doi.org/10.18653/v1/2021.acl-long.135 (Association for Computational Linguistics, 2021).
Sun, X. et al. On tracking dialogue state by inheriting slot values in mentioned slot pools. In Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence (IJCAI), 4369–4377 (2022).
Acknowledgements
We thank the original authors of the MultiWOZ 2.2 dataset and the Schema-Guided Dialogue (SGD) dataset for making these resources publicly available.
Funding
This work was supported by the Deanship of Scientific Research, Vice Presidency for Graduate Studies and Scientific Research, King Faisal University, Saudi Arabia Grant No. KFU260581.
Author information
Authors and Affiliations
Contributions
Abuzar Khan contributed to the conceptualization of the study, methodology design, and initial drafting of the manuscript. Fahad Masood assisted in data collection, experimental implementation, and result analysis. Abid Iqbal contributed to supervision, validation of results, and critical revision of the manuscript. Ahmad Junaid supported software development, simulations, and visualization of results. Saad Arif assisted in data preprocessing, performance evaluation, and literature review. Mohammed Al-Naeem contributed to formal analysis, resource provision, and manuscript review. Ghassan Husnain contributed to overall supervision, project administration, funding acquisition, and final manuscript approval. Ali Saeed Alzahrani contributed to conceptual guidance, critical proofreading, and technical refinement of the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethics Statement
This study does not involve human subjects, personal data or animal experiments and therefore does not require ethical approval.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Khan, A., Masood, F., Iqbal, A. et al. Evaluating routing stability and coordination in swarm-based multi-agent task-oriented dialogue systems. Sci Rep (2026). https://doi.org/10.1038/s41598-026-42158-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-42158-y