ResearchBenchmarksManipulation on OGBench puzzle-4x6-playFollow20Success RateORS-0.84.61015.4Apr 22, 2026Evaluation ResultsMethodMethodLinksSuccess RateORSseeds=8, test-time goa...seeds=8, test-time goals=52026.0420Go-Freshseeds=8, test-time goa...seeds=8, test-time goals=52026.0417GC-IQLseeds=8, test-time goa...seeds=8, test-time goals=52026.0412GC-IVLseeds=8, test-time goa...seeds=8, test-time goals=52026.0410CRLseeds=8, test-time goa...seeds=8, test-time goals=52026.044GCBCseeds=8, test-time goa...seeds=8, test-time goals=52026.040QRLseeds=8, test-time goa...seeds=8, test-time goals=52026.040