ResearchBenchmarksMathematical Reasoning on Olympiad (pass@1 and pass@32)Follow46.38Pass@1GRPO21.128827.684434.2440.7956May 12, 2026Evaluation ResultsMethodMethodLinksPass@1Pass@32GRPOBackbone=Qwen3-8B-BaseBackbone=Qwen3-8B-Base2026.0546.3859.33FG-ExPOBackbone=Qwen3-8B-BaseBackbone=Qwen3-8B-Base2026.0546.2560GRPOBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0530.2354FG-ExPOBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0529.3355.33BaseBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0522.144