ResearchBenchmarksValue Estimation on Suboptimal-Value-Bench ARX (retry segments)Value Estimation on Suboptimal-Value-Bench ARX (retry segments)Follow0.79Retry VOCWVM-0.3124-0.02620.260.5462Jun 23, 2026Evaluation resultsMethodMethodLinksRetry VOCWVMEmbodiment=ARXEmbodiment=ARX2026.060.79GVLEmbodiment=ARXEmbodiment=ARX2026.060.76TopRewardEmbodiment=ARXEmbodiment=ARX2026.06-0.19RobometerEmbodiment=ARXEmbodiment=ARX2026.06-0.27