Mathematical Reasoning on GSM8k (flexible_extract)
49.96Flexible Extract ScoreBaseline
Evaluation Results
| Method | Links | |
|---|---|---|
| BaselineModel=Llama 3.1-8B2026.02 | 49.96 | |
| TurboQuantModel=Llama 3.1-8B2026.02 | 48.9 | |
| InnerQBaseModel=Llama 3.1-8B2026.02 | 48.22 | |
| KIVISinkModel=Llama 3.1-8B2026.02 | 47.84 | |
| InnerQHybridModel=Llama 3.1-8B2026.02 | 47.31 | |
| InnerQSmallModel=Llama 3.1-8B2026.02 | 45.94 | |
| KIVIModel=Llama 3.1-8B2026.02 | 42.84 | |
| BaselineModel=Mistral 7B v0.32026.02 | 36.92 | |
| InnerQBaseModel=Mistral 7B v0.32026.02 | 36.47 | |
| TurboQuantModel=Mistral 7B v0.32026.02 | 34.72 | |
| InnerQHybridModel=Mistral 7B v0.32026.02 | 33.89 | |
| InnerQSmallModel=Mistral 7B v0.32026.02 | 33.89 | |
| KIVISinkModel=Mistral 7B v0.32026.02 | 33.81 | |
| KIVIModel=Mistral 7B v0.32026.02 | 32.3 | |
| InnerQBaseModel=Llama 3.2-3B2026.02 | 27.14 | |
| BaselineModel=Llama 3.2-3B2026.02 | 26.46 | |
| TurboQuantModel=Llama 3.2-3B2026.02 | 25.47 | |
| InnerQSmallModel=Llama 3.2-3B2026.02 | 24.03 | |
| KIVISinkModel=Llama 3.2-3B2026.02 | 23.35 | |
| InnerQHybridModel=Llama 3.2-3B2026.02 | 23.12 | |
| KIVIModel=Llama 3.2-3B2026.02 | 19.33 | |
| InnerQBaseModel=Llama 2-7B2026.02 | 13.42 | |
| KIVIModel=Llama 2-7B2026.02 | 13.12 | |
| BaselineModel=Llama 2-7B2026.02 | 13.04 | |
| TurboQuantModel=Llama 2-7B2026.02 | 12.59 | |
| KIVISinkModel=Llama 2-7B2026.02 | 11.9 | |
| InnerQHybridModel=Llama 2-7B2026.02 | 11.3 | |
| BaselineModel=Llama 3.2-1B2026.02 | 6.22 | |
| TurboQuantModel=Llama 3.2-1B2026.02 | 5.84 | |
| InnerQBaseModel=Llama 3.2-1B2026.02 | 5.53 | |
| InnerQHybridModel=Llama 3.2-1B2026.02 | 5.31 | |
| InnerQSmallModel=Llama 3.2-1B2026.02 | 4.63 | |
| KIVISinkModel=Llama 3.2-1B2026.02 | 4.02 | |
| KIVIModel=Llama 3.2-1B2026.02 | 2.81 | |
| InnerQSmallModel=Llama 2-7B2026.02 | 0.46 |