Knowledge-based numerical reasoning on NumGLUE Task 2 (test)
85.85AccuracyMultiTool-CoT
Evaluation Results
| Method | Links | |
|---|---|---|
| MultiTool-CoTExternal Tools=All (CAL, CRP, MML)2023.05 | 85.85 | |
| MultiTool-CoT (MML only)External Tools=Molar mass list only2023.05 | 69.23 | |
| MultiTool-CoT (CRP only)External Tools=Chemical reaction predictor only2023.05 | 64.31 | |
| MultiTool-CoT (CAL only)External Tools=Calculator only2023.05 | 62.77 | |
| Few-Shot+CoTPrompting=Few-Shot+CoT2023.05 | 57.85 | |
| Few-ShotPrompting=Few-Shot2023.05 | 42 | |
| Zero-Shot+CoTPrompting=Zero-Shot+CoT2023.05 | 32.62 | |
| Zero-ShotPrompting=Zero-Shot2023.05 | 1 |