ResearchBenchmarksMathematical Reasoning on AIME24 (Pass@16, Avg@16)Follow65.62Avg@16SENT-1.8427215.6716433.18650.70036Dec 1, 2025Dec 2, 2025Dec 3, 2025Dec 4, 2025Evaluation ResultsMethodMethodLinksAvg@16Pass@16SENTModel=Qwen3-14BModel=Qwen3-14B2025.1265.6286.67GRPOModel=Qwen3-14BModel=Qwen3-14B2025.1257.580ReProBackbone=Qwen3-8B, Bas...Backbone=Qwen3-8B, Base Training=GRPO2025.120.761—GRPOBackbone=Qwen3-8BBackbone=Qwen3-8B2025.120.756—OriginalBackbone=Qwen3-8BBackbone=Qwen3-8B2025.120.752—