List your company
ResearchFormerly WizWand

Multi-Agent Reinforcement Learning on Simple Spread

86.7Success RateQMIX + EMA
12.3431.64550.9570.255Jul 5, 2026
Updated 3mo ago

Evaluation results

MethodLinks
QMIX + EMARegime=Augmentative, Algorithm=QMIX, Stabilization method=EMA, Number of seeds=52026.0786.7
EMAalpha=0.22026.0786.7
Freeze-Schedule2026.0784.1
QMIX (Baseline)Regime=Augmentative, Algorithm=QMIX, Stabilization method=None, Number of seeds=52026.0774.4
Baseline QMIXshaping=no2026.0774.4
Dynamic LLMgranularity=per-episode2026.0715.2