ResearchBenchmarksMath reasoning on HMMT25 Nov.Follow15.83Mean@8MOPD9.3311.017512.70514.3925May 12, 2026Evaluation ResultsMethodMethodLinksMean@8Pass@8MOPDBase Model=Qwen3-4BBase Model=Qwen3-4B2026.0515.8322.47GRPOBase Model=Qwen3-4BBase Model=Qwen3-4B2026.0513.1619.53SDPOBase Model=Qwen3-4BBase Model=Qwen3-4B2026.0512.0118.49Qwen3-4Bstatus=base modelstatus=base model2026.059.5813.91