ResearchBenchmarksMathematical Reasoning on AMC23 (avg@32, Average)Follow91.9Avg@32DC-GRPO90.3490.74591.1591.555Jun 7, 2026Evaluation ResultsMethodMethodLinksAvg@32Average ScoreDC-GRPOBackbone=Qwen3-30B-A3B...Backbone=Qwen3-30B-A3B, Precision=BF162026.0691.971.8DC-GRPOBackbone=Qwen3-8B, Pre...Backbone=Qwen3-8B, Precision=BF162026.0691.871GRPOBackbone=Qwen3-30B-A3B...Backbone=Qwen3-30B-A3B, Precision=BF162026.0691.469GRPOBackbone=Qwen3-8B, Pre...Backbone=Qwen3-8B, Precision=BF162026.0690.467.3