ResearchBenchmarksMathematical reasoning on OlympiadBench (avg@4, Average)Follow74.5Avg@4 ScoreDC-GRPO69.92471.11272.373.488Jun 7, 2026Evaluation ResultsMethodMethodLinksAvg@4 ScoreAverage ScoreDC-GRPOBackbone=Qwen3-30B-A3B...Backbone=Qwen3-30B-A3B, Precision=BF162026.0674.571.8DC-GRPOBackbone=Qwen3-8B, Pre...Backbone=Qwen3-8B, Precision=BF162026.0674.471GRPOBackbone=Qwen3-30B-A3B...Backbone=Qwen3-30B-A3B, Precision=BF162026.0671.369GRPOBackbone=Qwen3-8B, Pre...Backbone=Qwen3-8B, Precision=BF162026.0670.167.3