ResearchBenchmarksSingle-life task completion on PointmassFollow61,000Average StepsQWALE55,51692,533129,550166,567Oct 17, 2022Evaluation ResultsMethodMethodLinksAverage StepsStandard ErrorSuccess RateMedian StepsQWALEmode=finetuningmode=finetuning2022.1061,00028,800801,700GAIL-smode=finetuningmode=finetuning2022.10100,90033,00050101,900GAIL-samode=finetuningmode=finetuning2022.10140,40030,30030200,000SAC-RNDmode=finetuningmode=finetuning2022.10184,10015,90010200,000SACmode=finetuningmode=finetuning2022.10198,1001,90010200,000