ResearchBenchmarksScience on GPQA D (Avg@4)Follow72.55Avg@4AIPO62.20264.888567.57570.2615May 8, 2026Evaluation ResultsMethodMethodLinksAvg@4AIPOModel=Qwen2.5-32B-Inst...Model=Qwen2.5-32B-Instruct2026.0572.55GRPOModel=Qwen2.5-32B-Inst...Model=Qwen2.5-32B-Instruct2026.0568.36SFTModel=Qwen2.5-32B-Inst...Model=Qwen2.5-32B-Instruct2026.0565.31OriginalModel=Qwen2.5-32B-Inst...Model=Qwen2.5-32B-Instruct2026.0562.6