ResearchBenchmarksReward Modeling on SummEval (test)Follow0.0444MSE (Table)FCPS0.04390.0472750.050650.054025May 31, 2026Evaluation ResultsMethodMethodLinksMSE (Table)MSE (Scalar)Delta MSE95% CI (Paired)FCPSnM=800nM=8002026.050.04440.04350.0009-0.0001FCPSnM=512nM=5122026.050.04480.04360.00110.0002FCPSnM=256nM=2562026.050.04510.046-0.0009-0.0022FCPSnM=128nM=1282026.050.04850.04740.0011-0.0006FCPSnM=64nM=642026.050.05280.05010.00260.0004FCPSnM=32nM=322026.050.05690.05110.00580.0032