Research map / Multi-agent systems and games / Cooperative multi-agent RL
ROMA: Multi-Agent Reinforcement Learning with Emergent Roles
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use ROMA’s emergent role-based learning instead of alternative MARL methods | MARL on the reported SMAC maps · StarCraft II micromanagement (SMAC) | win rate | Helped |
| Remove both role-learning objectives, L_I and L_D (ablation) | ROMA · MMM2 | win rate | Hurt |
| Add the specialization loss L_D (ablation) | ROMA without role-learning objectives · MMM2, 10z5b_vs_2s3z, and 10m_vs_11m | win rate | Helped |
| Add the identifiability loss L_I without L_D (ablation) | ROMA without role-learning objectives · MMM2, 10z5b_vs_2s3z, and 10m_vs_11m | win rate | Mixed |
| Remove the identifiability loss L_I while retaining L_D (ablation) | ROMA · SMAC ablation maps | training stability | Hurt |
| Remove the specialization loss L_D while retaining L_I (ablation) | ROMA · SMAC ablation maps | win rate | Hurt |
| Use independent agent parameters instead of shared parameters (ablation) | QMIX · SMAC | training speed | Hurt |
| Use ROMA’s dynamic sharing rather than fixed shared or independent parameters (ablation) | QMIX and QMIX-NPS · SMAC | team performance | Helped |
| Increase QMIX’s parameter count to approximately match ROMA (ablation) | QMIX · SMAC | team performance | Hurt |
| Use MAVEN instead of QMIX | MARL on SMAC · Most SMAC maps | win rate | Hurt |
| Compare trajectory dissimilarity across different unit types rather than the same unit type | ROMA dissimilarity model · MMM2 | trajectory dissimilarity: 0.0780\pm 0.0019 -> 0.9556\pm 0.0009 | Helped |
Closest papers by meaning
- R3DM: Enabling Role Discovery and Diversity Through Dynamics Models in Multi-agent Reinforcement Learning (ICML 2025 · similarity 0.91)
- Attention-Guided Contrastive Role Representations for Multi-agent Reinforcement Learning (ICLR 2024 · similarity 0.89)
- ComaDICE: Offline Cooperative Multi-Agent Reinforcement Learning with Stationary Distribution Shift Regularization (ICLR 2025 · similarity 0.86)
- Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.86)
- Foundations of Multi-Agent Learning in Dynamic Environments: Where Reinforcement Learning Meets Strategic Decision-Making (AAAI 2025 · similarity 0.85)
- JaxMARL: Multi-Agent RL Environments and Algorithms in JAX (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.