ResearchBenchmarksLogical Reasoning on BBH 3-shot chain-of-thoughtFollow61.35EMVAR10.826823.943437.0650.1766Feb 16, 2025Evaluation ResultsMethodMethodLinksEMVARBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2025.0261.35ALoLBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2025.0261.14DPOBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2025.0253.06DPOBase Model=Llama2-7BBase Model=Llama2-7B2025.0239.1ALoLBase Model=Llama2-7BBase Model=Llama2-7B2025.0238.16VARBase Model=Llama2-7BBase Model=Llama2-7B2025.0237.56BaseBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2025.0229.3BaseBase Model=Llama2-7BBase Model=Llama2-7B2025.0212.77