ResearchBenchmarksMathematical Problem Solving on MATH500 (Pass@32)Follow0.668Avg@32MASPO0.664880.665690.66650.66731Feb 19, 2026Evaluation ResultsMethodMethodLinksAvg@32Pass@32MASPOBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-Qwen-14B2026.020.6680.733GRPOBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-Qwen-14B2026.020.6650.704