Research map / Multi-agent systems and games / Cooperative multi-agent RL
RMIX: Learning Risk-Sensitive Policies for Cooperative Reinforcement Learning Agents
Wei Qiu, Xinrun Wang, Runsheng Yu, Rundong Wang, Xu He, Bo An, Svetlana Obraztsova, Zinovi Rabinovich
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| learn individual CVaR risk-sensitive policies instead of expected Q values | RMIX · SMAC asymmetric homogeneous maps | test win rate | Improved |
| use RMIX versus MAVEN for exploration | RMIX · SMAC corridor | steps to converge: 7 million -> 5 million | Improved |
| learn individual return distributions instead of a centralized distribution | RMIX vs QR-MIX · 6 SMAC scenarios | test win rate | Improved |
| use CVaR over individual distributions versus likelihood hysteretic IQN | RMIX vs LH-IQN · SMAC scenarios | test win rate | Improved |
| fix static risk level alpha in {0.1,0.3,0.5,0.7,1.0} | RMIX · 1c3s5z and 5m_vs_6m | test win rate | Improved |
| add risk level predictor instead of static alpha (ablation) | RMIX · 1c3s5z and 5m_vs_6m | convergence speed | Improved |
| apply CVaR framework to VDN additive mixer (RDN) | RDN · 10m_vs_11m | test win rate | Improved |
| apply CVaR framework to VDN additive mixer (RDN) | RDN · 1c3s5z and 8m_vs_9m | convergence speed | Improved |
| centralized distribution (QR-MIX) versus expected Q baselines | QR-MIX · 1c3s5z, MMM2, 5m_vs_6m | test win rate | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MAVEN — MAVEN: Multi-Agent Variational Exploration (NeurIPS 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Disentangling Sources of Risk for Distributional Multi-Agent Reinforcement Learning (ICML 2022 · similarity 0.90)
- Provably Efficient Iterated CVaR Reinforcement Learning with Function Approximation and Human Feedback (ICLR 2024 · similarity 0.88)
- Empirical Study on Robustness and Resilience in Cooperative Multi-Agent Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Learning Joint Behaviors with Large Variations (AAAI 2025 · similarity 0.88)
- Scalable Constrained Policy Optimization for Safe Multi-agent Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Risk-Sensitive Reward-Free Reinforcement Learning with CVaR (ICML 2024 · similarity 0.87)
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.
