Research map / Multi-agent systems and games / Cooperative multi-agent RL
Effective and Stable Role-Based Multi-Agent Collaboration by Structural Information Principles
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 SIRD hierarchical role discovery in place of the strongest reported baseline | SR-MARL · SMAC easy maps | average test win rate: ↑0.17 | Helped |
| Use SIRD hierarchical role discovery in place of the strongest reported baseline | SR-MARL · SMAC hard maps | average test win rate: ↑6.08 | Helped |
| Use SIRD hierarchical role discovery in place of the strongest reported baseline | SR-MARL · SMAC super hard maps | average test win rate: ↑3.24 | Helped |
| Use SIRD hierarchical role discovery | SR-MARL · SMAC easy maps | test win rate deviation: ↓16.67 | Helped |
| Use SIRD hierarchical role discovery | SR-MARL · SMAC hard maps | test win rate deviation: ↓30.80 | Helped |
| Use SIRD hierarchical role discovery | SR-MARL · SMAC super hard maps | test win rate deviation: ↓66.30 | Helped |
| Add action-graph structuralization instead of discovering roles with K-Means on the joint action space (ablation) | SR-MARL · 2c_vs_64zg | average test win rate | Helped |
| Add action-graph structuralization instead of discovering roles with K-Means on the joint action space (ablation) | SR-MARL · 2c_vs_64zg | test win rate deviation | Helped |
| Add sparsification before optimizing the encoding tree (ablation) | SR-MARL · 2c_vs_64zg | learning process | Helped |
| Add sparsification before optimizing the encoding tree (ablation) | SR-MARL · 2c_vs_64zg | performance advantages | No clear effect |
| Increase maximal encoding-tree height from 2 to 3 (ablation) | SR-MARL · MMM2 | effectiveness, stability, and efficiency | Helped |
| Increase maximal encoding-tree height from 2 to 3 (ablation) | SR-MARL · 2c_vs_64zg | effectiveness, stability, and efficiency | Hurt |
Closest papers by meaning
- RODE: Learning Roles to Decompose Multi-Agent Tasks (ICLR 2021 · similarity 0.91)
- GRDC: A Unified Graph-Driven Framework for Role Discovery and Communication in Multi-Agent Reinforcement Learning (AAAI 2026 · similarity 0.91)
- R3DM: Enabling Role Discovery and Diversity Through Dynamics Models in Multi-agent Reinforcement Learning (ICML 2025 · similarity 0.90)
- Attention-Guided Contrastive Role Representations for Multi-agent Reinforcement Learning (ICLR 2024 · similarity 0.88)
- Learning Generalizable Skills from Offline Multi-Task Data for Multi-Agent Cooperation (ICLR 2025 · similarity 0.85)
- Sequential Order Adjustment of Action Decisions for Multi-Agent Transformer (Student Abstract) (AAAI 2025 · 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.