ResearchBenchmarksMathematical Reasoning on AIME 25 (Accuracy and Token Count)Follow24.38AIME 25 AccuracyMUR9.643213.469117.29521.1209Jul 20, 2025Evaluation ResultsMethodMethodLinksAIME 25 AccuracyAverage TokensMURBackbone=Qwen3-8B, Sea...Backbone=Qwen3-8B, Search Strategy=phi-Decoding2025.0724.3822,296MURBackbone=Qwen3-4B, Sea...Backbone=Qwen3-4B, Search Strategy=LLM as a critic2025.0720.213,113MURBackbone=Qwen3-1.7B, S...Backbone=Qwen3-1.7B, Search Strategy=LLM as a critic2025.0710.214,011