STEM Task Evaluation on MMLU Biology
0.8819AccuracyRe-TASK (Lite)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Re-TASK (Lite)Model=Llama3-8B2024.08 | 0.8819 | 18.29 | |
| Re-TASK (Lite)Model=Qwen1.5-7B2024.08 | 0.8125 | 18.29 | |
| One-shot CoTModel=Llama3-8B2024.08 | 0.7986 | 5.09 | |
| Re-TASK (Lite)Model=Mistral-7B2024.08 | 0.7917 | 18.29 | |
| Zero-shot CoT+SCModel=Llama3-8B2024.08 | 0.7847 | 2.55 | |
| Zero-shot CoTModel=Llama3-8B2024.08 | 0.7639 | — | |
| Plan-and-SolveModel=Llama3-8B2024.08 | 0.7361 | -1.16 | |
| One-shot CoTModel=Mistral-7B2024.08 | 0.6875 | 5.09 | |
| Zero-shot CoT+SCModel=Qwen1.5-7B2024.08 | 0.625 | 2.55 | |
| Plan-and-SolveModel=Qwen1.5-7B2024.08 | 0.6111 | -1.16 | |
| Zero-shot CoT+SCModel=Mistral-7B2024.08 | 0.6042 | 2.55 | |
| One-shot CoTModel=Qwen1.5-7B2024.08 | 0.6042 | 5.09 | |
| Zero-shot CoTModel=Qwen1.5-7B2024.08 | 0.5972 | — | |
| Zero-shot CoTModel=Mistral-7B2024.08 | 0.5764 | — | |
| Plan-and-SolveModel=Mistral-7B2024.08 | 0.5556 | -1.16 | |
| Step-BackModel=Qwen1.5-7B2024.08 | 0.5347 | -15.28 | |
| Step-BackModel=Mistral-7B2024.08 | 0.5069 | -15.28 | |
| Step-BackModel=Llama3-8B2024.08 | 0.4375 | -15.28 |