Disentangling Sources of Risk for Distributional Multi-Agent Reinforcement Learning

Benchmark Model Rank Results
smac-on-smac-def-armored-parallelDRIMA–Median Win Rate: 60.0
smac-on-smac-def-armored-sequentialDRIMA–Median Win Rate: 100
smac-on-smac-def-infantry-parallelDRIMA–Median Win Rate: 100.0
smac-on-smac-def-infantry-sequentialDRIMA–Median Win Rate: 100
smac-on-smac-def-outnumbered-parallelDRIMA–Median Win Rate: 70.0
smac-on-smac-def-outnumbered-sequentialDRIMA–Median Win Rate: 100
smac-on-smac-off-complicated-parallelDRIMA–Median Win Rate: 100
smac-on-smac-off-complicated-sequentialDRIMA–Median Win Rate: 96.9
smac-on-smac-off-distant-parallelDRIMA–Median Win Rate: 95.0
smac-on-smac-off-distant-sequentialDRIMA–Median Win Rate: 100
smac-on-smac-off-hard-parallelDRIMA–Median Win Rate: 80.0
smac-on-smac-off-hard-sequentialDRIMA–Median Win Rate: 93.8
smac-on-smac-off-near-parallelDRIMA–Median Win Rate: 95.0
smac-on-smac-off-near-sequentialDRIMA–Median Win Rate: 93.8
smac-on-smac-off-superhard-parallelDRIMA–Median Win Rate: 0.0
smac-on-smac-off-superhard-sequentialDRIMA–Median Win Rate: 15.6