ResearchBenchmarksMathematical Reasoning on AIME24 (Average Accuracy)Follow23.5Average AccuracyGRPO11.0214.2617.520.74Dec 1, 2025Evaluation ResultsMethodMethodLinksAverage AccuracyGRPOBackbone=Qwen3-4B-Base...Backbone=Qwen3-4B-Base, Sample Count=Avg@16, Evaluation Domain=In-domain2025.1223.5ReProBackbone=Qwen3-4B-Base...Backbone=Qwen3-4B-Base, Sample Count=Avg@16, Evaluation Domain=In-domain, Base Method=GRPO2025.1221OriginalBackbone=Qwen3-4B-Base...Backbone=Qwen3-4B-Base, Sample Count=Avg@16, Evaluation Domain=In-domain2025.1211.5