List your company
ResearchFormerly WizWand

STEM Task Evaluation on MMLU Math

51.81AccuracyRe-TASK (Lite)
18.654827.262435.8744.4776Aug 13, 2024
Updated 8mo ago

Evaluation results

MethodLinks
Re-TASK (Lite)Model=Llama3-8B2024.0851.817.61
One-shot CoTModel=Llama3-8B2024.0849.422.51
Zero-shot CoT+SCModel=Llama3-8B2024.0848.194.23
Step-BackModel=Llama3-8B2024.0845.65-0.6
Re-TASK (Lite)Model=Qwen1.5-7B2024.0843.847.61
Zero-shot CoT+SCModel=Qwen1.5-7B2024.0841.674.23
Zero-shot CoTModel=Llama3-8B2024.0840.58—
Plan-and-SolveModel=Llama3-8B2024.0840.58-2.3
Zero-shot CoTModel=Qwen1.5-7B2024.0836.96—
One-shot CoTModel=Qwen1.5-7B2024.0836.522.51
Step-BackModel=Mistral-7B2024.0834.42-0.6
Plan-and-SolveModel=Mistral-7B2024.0830.43-2.3
Re-TASK (Lite)Model=Mistral-7B2024.0828.997.61
Zero-shot CoT+SCModel=Mistral-7B2024.0824.644.23
Zero-shot CoTModel=Mistral-7B2024.0824.28—
Plan-and-SolveModel=Qwen1.5-7B2024.0823.91-2.3
One-shot CoTModel=Mistral-7B2024.0823.412.51
Step-BackModel=Qwen1.5-7B2024.0819.93-0.6