ResearchBenchmarksOutcome Reasoning on ArithmeticFollow87.8M' F1 MeanGPT-557.95265.70173.4581.199May 17, 2025Evaluation ResultsMethodMethodLinksM' F1 MeanY' F1 MeanGPT-5Model=GPT-5Model=GPT-52025.0587.882.7GPT-o4Model=GPT-o4Model=GPT-o42025.0585.880.9Llama4-MModel=Llama4-MModel=Llama4-M2025.0576.370.6DeepSeekModel=DeepSeekModel=DeepSeek2025.057467.5Gemini2.5Model=Gemini2.5Model=Gemini2.52025.0572.165.4Qwen3Model=Qwen3Model=Qwen32025.0569.863.2Llama4-SModel=Llama4-SModel=Llama4-S2025.0559.152.4