Multi-Agent Reinforcement Learning on Battle Medium N=10,000
86.1Normalized ReturnMF-Diffuser
Evaluation Results
| Method | Links | |
|---|---|---|
| MF-Diffuser2026.05 | 86.1 | |
| MF-CDMs-RL2026.05 | 71.2 | |
| Oryx2026.05 | 70 | |
| MFQ-Offline2026.05 | 63.8 | |
| DoF2026.05 | 52.5 | |
| Indep. Diffuser2026.05 | 47.2 | |
| MA-TD3+BC2026.05 | 18.7 | |
| OMAR2026.05 | 10.2 | |
| MADiffattention=chunked-attention (max 12-agent windows)2026.05 | 6.1 | |
| Joint Diffuserattention=chunked-attention (block size 32)2026.05 | 3.4 |