ResearchBenchmarksMathematical Reasoning on BRUMO 2025 (Majority@2048)Follow23.3Majority@2048CurveRL12.915.618.321May 23, 2026Evaluation ResultsMethodMethodLinksMajority@2048CurveRLBackbone=Qwen3-1.7B-BaseBackbone=Qwen3-1.7B-Base2026.0523.3GRPOBackbone=Qwen3-1.7B-BaseBackbone=Qwen3-1.7B-Base2026.0516.7MaxRLBackbone=Qwen3-1.7B-BaseBackbone=Qwen3-1.7B-Base2026.0516.7BaseBackbone=Qwen3-1.7B-BaseBackbone=Qwen3-1.7B-Base2026.0513.3