From Traits to Roles: Consensus-Guided Composition of Orthogonal Experts for Cooperative MARL

Zhou, Yewei; Zhang, Bin; Zhou, Ying; Ge, Xuri; Li, Dapeng; Mao, Hangyu; Ren, Pengjie; Xu, Zhiwei

doi:10.24963/ijcai.2026/55

From Traits to Roles: Consensus-Guided Composition of Orthogonal Experts for Cooperative MARL

Yewei Zhou, Bin Zhang, Ying Zhou, Xuri Ge, Dapeng Li, Hangyu Mao, Pengjie Ren, Zhiwei Xu

Download BibTeX Cite

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence. Main Track. Pages 481-489.

https://doi.org/10.24963/ijcai.2026/55

Keywords

Agent-based and Multi-agent Systems: Multi-agent learning

Abstract

Parameter sharing is a central design choice in cooperative multi-agent reinforcement learning, yet it fundamentally conflicts with the need for role specialization in heterogeneous cooperative environments. Existing role-based methods typically learn monolithic role representations, which often suffer from gradient interference and fail to capture the compositional structure of complex behaviors. Inspired by Trait Theory, we propose DEcompose and COnstruct Roles (DECOR), a framework that models agent roles as dynamic compositions of orthogonal behavioral traits. DECOR introduces an orthogonal Mixture-of-Experts architecture to decompose behaviors into independent traits, mitigating destructive gradient interference under parameter sharing, and a group-consensus guided mechanism to extract team-level tactical intents that guide role composition.Experiments on multiple benchmarks demonstrate that DECOR consistently improves sample efficiency and overall performance over existing related methods.

Yewei Zhou, Bin Zhang, Ying Zhou, Xuri Ge, Dapeng Li, Hangyu Mao, Pengjie Ren, and Zhiwei Xu. 2026. From Traits to Roles: Consensus-Guided Composition of Orthogonal Experts for Cooperative MARL. In Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26. IJCAI, 481–489. https://doi.org/10.24963/ijcai.2026/55

@inproceedings{ijcai2026p55,
  title     = {From Traits to Roles: Consensus-Guided Composition of Orthogonal Experts for Cooperative MARL},
  author    = {Zhou, Yewei and Zhang, Bin and Zhou, Ying and Ge, Xuri and Li, Dapeng and Mao, Hangyu and Ren, Pengjie and Xu, Zhiwei},
  booktitle = {Proceedings of the Thirty-Fifth International Joint Conference on
               Artificial Intelligence, {IJCAI-26}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {Diego Calvanese},
  pages     = {481--489},
  year      = {2026},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2026/55},
  url       = {https://doi.org/10.24963/ijcai.2026/55},
}

TY  - CPAPER
TI  - From Traits to Roles: Consensus-Guided Composition of Orthogonal Experts for Cooperative MARL
AU  - Zhou, Yewei
AU  - Zhang, Bin
AU  - Zhou, Ying
AU  - Ge, Xuri
AU  - Li, Dapeng
AU  - Mao, Hangyu
AU  - Ren, Pengjie
AU  - Xu, Zhiwei
T2  - Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2026
SP  - 481
EP  - 489
DO  - 10.24963/ijcai.2026/55
UR  - https://doi.org/10.24963/ijcai.2026/55
ER  -