On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agents

Jen-Tse Huang, Jiaxu Zhou, Tailin Jin, Xuhui Zhou, Zixi Chen, Wenxuan Wang, Youliang Yuan, Michael Lyu, Maarten Sap
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:26202-26226, 2025.

Abstract

Large language model-based multi-agent systems have shown great abilities across various tasks due to the collaboration of expert agents, each focusing on a specific domain. However, the impact of clumsy or even malicious agents—those who frequently make errors in their tasks—on the overall performance of the system remains underexplored. This paper investigates: (1) What is the resilience of various system structures (e.g., A$\rightarrow$B$\rightarrow$C, A$\leftrightarrow$B$\leftrightarrow$C) under faulty agents, on different downstream tasks? (2) How can we increase system resilience to defend against these agents? To simulate faulty agents, we propose two approaches—AutoTransform and AutoInject—which introduce mistakes into the agents’ responses. Experiments on four downstream tasks using six systems show that the "hierarchical" structure, i.e., A$\rightarrow$(B$\leftrightarrow$C), exhibits superior resilience with the lowest performance drop of 5.5%, compared to 10.5% and 23.7% of other two structures. To further improve resilience, we introduce (1) Challenger, that introduces a mechanism for each agent to challenge others’ outputs, and (2) Inspector, an additional agent to review and correct messages, recovering up to 96.4% errors made by faulty agents. Our code and data are available at https://github.com/CUHK-ARISE/MAS-Resilience.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-huang25ay, title = {On the Resilience of {LLM}-Based Multi-Agent Collaboration with Faulty Agents}, author = {Huang, Jen-Tse and Zhou, Jiaxu and Jin, Tailin and Zhou, Xuhui and Chen, Zixi and Wang, Wenxuan and Yuan, Youliang and Lyu, Michael and Sap, Maarten}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {26202--26226}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/huang25ay/huang25ay.pdf}, url = {https://proceedings.mlr.press/v267/huang25ay.html}, abstract = {Large language model-based multi-agent systems have shown great abilities across various tasks due to the collaboration of expert agents, each focusing on a specific domain. However, the impact of clumsy or even malicious agents—those who frequently make errors in their tasks—on the overall performance of the system remains underexplored. This paper investigates: (1) What is the resilience of various system structures (e.g., A$\rightarrow$B$\rightarrow$C, A$\leftrightarrow$B$\leftrightarrow$C) under faulty agents, on different downstream tasks? (2) How can we increase system resilience to defend against these agents? To simulate faulty agents, we propose two approaches—AutoTransform and AutoInject—which introduce mistakes into the agents’ responses. Experiments on four downstream tasks using six systems show that the "hierarchical" structure, i.e., A$\rightarrow$(B$\leftrightarrow$C), exhibits superior resilience with the lowest performance drop of 5.5%, compared to 10.5% and 23.7% of other two structures. To further improve resilience, we introduce (1) Challenger, that introduces a mechanism for each agent to challenge others’ outputs, and (2) Inspector, an additional agent to review and correct messages, recovering up to 96.4% errors made by faulty agents. Our code and data are available at https://github.com/CUHK-ARISE/MAS-Resilience.} }
Endnote
%0 Conference Paper %T On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agents %A Jen-Tse Huang %A Jiaxu Zhou %A Tailin Jin %A Xuhui Zhou %A Zixi Chen %A Wenxuan Wang %A Youliang Yuan %A Michael Lyu %A Maarten Sap %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-huang25ay %I PMLR %P 26202--26226 %U https://proceedings.mlr.press/v267/huang25ay.html %V 267 %X Large language model-based multi-agent systems have shown great abilities across various tasks due to the collaboration of expert agents, each focusing on a specific domain. However, the impact of clumsy or even malicious agents—those who frequently make errors in their tasks—on the overall performance of the system remains underexplored. This paper investigates: (1) What is the resilience of various system structures (e.g., A$\rightarrow$B$\rightarrow$C, A$\leftrightarrow$B$\leftrightarrow$C) under faulty agents, on different downstream tasks? (2) How can we increase system resilience to defend against these agents? To simulate faulty agents, we propose two approaches—AutoTransform and AutoInject—which introduce mistakes into the agents’ responses. Experiments on four downstream tasks using six systems show that the "hierarchical" structure, i.e., A$\rightarrow$(B$\leftrightarrow$C), exhibits superior resilience with the lowest performance drop of 5.5%, compared to 10.5% and 23.7% of other two structures. To further improve resilience, we introduce (1) Challenger, that introduces a mechanism for each agent to challenge others’ outputs, and (2) Inspector, an additional agent to review and correct messages, recovering up to 96.4% errors made by faulty agents. Our code and data are available at https://github.com/CUHK-ARISE/MAS-Resilience.
APA
Huang, J., Zhou, J., Jin, T., Zhou, X., Chen, Z., Wang, W., Yuan, Y., Lyu, M. & Sap, M.. (2025). On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agents. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:26202-26226 Available from https://proceedings.mlr.press/v267/huang25ay.html.

Related Material