ResearchBenchmarksMathematical Reasoning on AIME 25 (Acc@16)Follow55.2Acc@16AMR-SD46.15248.50150.8553.199May 18, 2026Evaluation ResultsMethodMethodLinksAcc@16AMR-SDBackbone=Qwen3-8B, Thi...Backbone=Qwen3-8B, Thinking Mode=true2026.0555.2GRPOBackbone=Qwen3-8B, Thi...Backbone=Qwen3-8B, Thinking Mode=true2026.0553.1RLSDBackbone=Qwen3-8B, Thi...Backbone=Qwen3-8B, Thinking Mode=true2026.0547.3Qwen3-8BThinking Mode=trueThinking Mode=true2026.0546.5