ResearchBenchmarksOutcome Reasoning on CVQA CountFollow0.792F1 Mean (M')GPT-50.5060.580250.65450.72875May 17, 2025Evaluation ResultsMethodMethodLinksF1 Mean (M')F1 Mean (Y')GPT-5Model=GPT-5Model=GPT-52025.050.7920.72GPT-o4Model=GPT-o4Model=GPT-o42025.050.7620.693Llama4-MModel=Llama4-MModel=Llama4-M2025.050.6750.61DeepSeekModel=DeepSeekModel=DeepSeek2025.050.6350.569Qwen3Model=Qwen3Model=Qwen32025.050.6220.557Gemini2.5Model=Gemini2.5Model=Gemini2.52025.050.6010.538Llama4-SModel=Llama4-SModel=Llama4-S2025.050.5170.452