ResearchBenchmarksReward Modeling on Arena100K (test)Follow0.2311Table MSEFCPS0.2308120.2327560.23470.236644May 31, 2026Evaluation ResultsMethodMethodLinksTable MSEScalar MSEDelta MSEPaired 95% CI Lower BoundFCPSnM=400nM=4002026.050.23110.2321-0.001-0.0045FCPSnM=100nM=1002026.050.23260.2349-0.0023-0.0056FCPSnM=200nM=2002026.050.23280.2331-0.0003-0.0035FCPSnM=50nM=502026.050.23830.23720.0011-0.0033