ResearchBenchmarksVisual Question Answering on CONTEXTCLARIFYVisual Question Answering on CONTEXTCLARIFYFollow47.4VQA AccuracyCoA9.02418.98728.9538.913Jan 23, 2026Evaluation resultsMethodMethodLinksVQA AccuracyCoAModel=Qwen2.5-VL, Mode...Model=Qwen2.5-VL, Model size=7B2026.0147.4CoAModel=InternVL3, Model...Model=InternVL3, Model size=2B2026.0136.8PromptModel=Qwen2.5-VL, Mode...Model=Qwen2.5-VL, Model size=7B2026.0131.6CoAModel=Qwen2.5-VL, Mode...Model=Qwen2.5-VL, Model size=3B2026.0131.6PromptModel=Qwen2.5-VL, Mode...Model=Qwen2.5-VL, Model size=3B2026.0123.6PromptModel=InternVL3, Model...Model=InternVL3, Model size=2B2026.0110.5