Research map / Multi-agent systems and games / Cooperative multi-agent RL
Self-Organized Group for Cooperative Multi-agent Reinforcement Learning
Jianzhun Shao, Zhiqiang Lou, Hongchang Zhang, Yuhang Jiang, Shuncheng He, Xiangyang Ji
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 |
|---|---|---|---|
| use L_FP + L_CEB message summarizer losses together (ablation) | SOG · Predator-Prey complicated scenarios | average return: 7.2 -> 9.98 | Improved |
| add entropy regularizer to message loss (ablation) | SOG · Predator-Prey complicated scenarios | average return: 9.98 -> 7.31 | Hurt |
| add entropy regularizer to reduce variance (ablation) | SOG · Predator-Prey | return std: 1.15 -> 0.08 | Improved |
| number of groups 1 to 2 (ablation) | SOG · Predator-Prey | test return: 7.43 -> 9.90 | Improved |
| number of groups 2 to 4 (ablation) | SOG · Predator-Prey | test return: 9.90 -> 7.37 | Hurt |
| message dimension 10 to 3 (ablation) | SOG · Predator-Prey | test return: 10.77 -> 9.90 | No clear effect |
| message dimension 3 to 1 (ablation) | SOG · Predator-Prey | test return: 9.90 -> 6.30 | Hurt |
| communication interval T 4 to 2 (ablation) | SOG · Predator-Prey | test return: 9.90 -> 10 | Improved |
| communication interval T 4 to 10 (ablation) | SOG · Predator-Prey | test return: 9.90 -> 7.5 | Hurt |
| SOG vs EMP on complicated evaluation scenarios | SOG · Predator-Prey | average return: 9.37 -> 9.98 | Improved |
| PG-based conductor election instead of random (ablation) | SOG · StarCraft 3-8sz | test winning rate | Improved |
| DPP/PG conductor election vs random on team-size changes (ablation) | SOG · Resource Collection | average episode return | No clear effect |
Closest papers by meaning
- Automatic Grouping for Efficient Cooperative Multi-Agent Reinforcement Learning (NeurIPS 2023 · similarity 0.89)
- GTDE: Grouped Training with Decentralized Execution for Multi-agent Actor-Critic (AAAI 2025 · similarity 0.86)
- Intrinsic Action Tendency Consistency for Cooperative Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.86)
- Bridging Training and Execution via Dynamic Directed Graph-Based Communication in Cooperative Multi-Agent Systems (AAAI 2025 · similarity 0.85)
- Efficient Communication in Multi-Agent Reinforcement Learning with Implicit Consensus Generation (AAAI 2025 · similarity 0.85)
- Beyond Training Time, Test-Time Coordination is Essential for Cooperative MARL (NeurIPS 2026 · 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.
