ResearchBenchmarksContinuous control locomotion on MuJoCo Ant v3 (train)Follow762Avg Performance (1M Steps)GePPO536.32594.91653.5712.09Oct 29, 2021Evaluation ResultsMethodMethodLinksAvg Performance (1M Steps)Final PerformanceSteps to Final Performance (M)Percent ImprovementGePPOclipping parameter=0.1...clipping parameter=0.1, value function estimator=V-trace2021.107621,5760.7940PPOclipping parameter=0.2...clipping parameter=0.2, batch size=20482021.105451,2271—