Multi-Agent Reinforcement Learning on SMAC 3m
99.9Win RateQMIX + EMA
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| QMIX + EMARegime=Supplementary, Algorithm=QMIX, Stabilization method=EMA, Number of seeds=52026.07 | 99.9 | — | — | — | — | — | |
| EMAalpha=0.22026.07 | 99.9 | — | — | — | — | — | |
| Freeze-Schedule2026.07 | 99.8 | — | — | — | — | — | |
| QMIX (Baseline)Regime=Supplementary, Algorithm=QMIX, Stabilization method=None, Number of seeds=52026.07 | 98.8 | — | — | — | — | — | |
| Baseline QMIXshaping=no2026.07 | 98.8 | — | — | — | — | — | |
| Dynamic LLMgranularity=per-episode2026.07 | 96.4 | — | — | — | — | — | |
| IBAL2026.05 | 67.1 | — | 25 | 76 | 83.8 | 1.7 | |
| FGSM2026.05 | 39.3 | — | 1 | 70.7 | 81.2 | 4.3 | |
| ATLA2026.05 | 39 | — | 0 | 77.3 | 78.1 | 4.7 | |
| ERNIE2026.05 | 35.9 | — | 3.1 | 62.7 | 74 | 1.3 | |
| WALL2026.05 | 35.4 | — | 0 | 70 | 67.7 | 4 | |
| ROMANCE2026.05 | 34.7 | — | 0 | 64.7 | 70.8 | 3.3 | |
| Rand-Obs2026.05 | 34.1 | — | 0 | 62 | 70.3 | 6 | |
| Vanilla QMIX2026.05 | 30.5 | — | 0 | 56.3 | 62 | 1.7 | |
| Rand-Act2026.05 | 27.1 | — | 0 | 41.7 | 66.7 | 0 | |
| MAT-DecDifficulty=Easy, Sight Range=92026.01 | 1 | 1.1 | — | — | — | — | |
| MAPPODifficulty=Easy, Sight Range=92026.01 | 1 | 0.4 | — | — | — | — | |
| HAPPODifficulty=Easy, Sight Range=92026.01 | 1 | 1.2 | — | — | — | — | |
| DG-MAPPODifficulty=Easy, Sight Range=92026.01 | 1 | 1.4 | — | — | — | — |