Multi-Agent Reinforcement Learning on Battle Med-Replay N=10,000
80.5Normalized ReturnMF-Diffuser
Evaluation Results
| Method | Links | |
|---|---|---|
| MF-Diffuser2026.05 | 80.5 | |
| MF-CDMs-RL2026.05 | 66.3 | |
| Oryx2026.05 | 64.8 | |
| MFQ-Offline2026.05 | 57.5 | |
| DoF2026.05 | 46.8 | |
| Indep. Diffuser2026.05 | 41.8 | |
| MA-TD3+BC2026.05 | 13.4 | |
| OMAR2026.05 | 6.8 | |
| MADiffattention=chunked-attention (max 12-agent windows)2026.05 | 4 | |
| Joint Diffuserattention=chunked-attention (block size 32)2026.05 | 2.1 |