List your company

Mathematical Reasoning on Math500, OlympiadBench, Minerva, AIME (24/25), and AMC

85.6Math500 AccuracyDMPO
62.7268.6674.680.54May 19, 2026

Evaluation results

MethodLinks
DMPO2026.0585.648.441.220.21556.344.5
GRPO2026.0583.844.939.717.814.154.742.5
GSPO2026.0583.244.739.318.812.653.942.1
FlowRL2026.0582.44437.61611.150.240.2
ClipCov2026.0581.446.738.619.113.854.542.3
GPG2026.0579.642.740.416.310.648.839.7
Backbone2026.0563.628.321.7146.945.630