ResearchBenchmarksManipulation on Adroit Pen-cloned (human-labeled)Follow89PerformanceFB-PbRL11.41631.55851.771.842May 31, 2026Evaluation ResultsMethodMethodLinksPerformanceFB-PbRLfine-tuning=truefine-tuning=true2026.0589DPPO2026.0575.1LiRE2026.0514.4