ResearchBenchmarksMulti-agent Reinforcement Learning on HalfCheetah 2x3Follow5,176Mean Episode RewardHPA2,892.163,485.084,0784,670.92May 8, 2026Evaluation ResultsMethodMethodLinksMean Episode RewardMax Episode RewardHPA2026.055,1765,432HAPPO2026.054,5025,002STEP2026.054,3944,899HATRPO2026.053,8534,238MAPPO2026.052,9803,002