ResearchBenchmarksOffline Reinforcement Learning on OGBench puzzle-4x4-play (5 tasks)Follow30Success RatePCBF6.0812.2918.524.71May 7, 2026Evaluation ResultsMethodMethodLinksSuccess RateNormalized ReturnPCBFSeeds=8Seeds=82026.0530—FloQSeeds=8Seeds=82026.0528—IQNSeeds=8Seeds=82026.0527—Value FlowsSeeds=8Seeds=82026.0527—CODACSeeds=8Seeds=82026.0520—FQLSeeds=8Seeds=82026.0517—IQLSeeds=8Seeds=82026.057—