ResearchBenchmarksMultimodal Understanding and Reasoning on MMMU-ProFollow44.7AccuracyPTA-GRPO36.58838.69440.842.906Oct 2, 2025Evaluation ResultsMethodMethodLinksAccuracyPTA-GRPOBase Model=Qwen2.5-7B-VLBase Model=Qwen2.5-7B-VL2025.1044.7SRPOBase Model=Qwen2.5-7B-VLBase Model=Qwen2.5-7B-VL2025.1042.3BaseBase Model=Qwen2.5-7B-VLBase Model=Qwen2.5-7B-VL2025.1036.9