Question Answering on TriviaQA LongBench
91.7F1 ScoreInnerQHybrid
Evaluation results
| Method | Links | |
|---|---|---|
| InnerQHybridModel=Llama 3.1-8B2026.02 | 91.7 | |
| InnerQSmallModel=Llama 3.1-8B2026.02 | 91.7 | |
| BaselineModel=Llama 3.1-8B2026.02 | 91.64 | |
| InnerQBaseModel=Llama 3.1-8B2026.02 | 91.51 | |
| KIVISinkModel=Llama 3.1-8B2026.02 | 91.33 | |
| KIVIModel=Llama 3.1-8B2026.02 | 90.53 | |
| InnerQHybridModel=Llama 2-13B2026.02 | 88.27 | |
| KIVIModel=Llama 2-13B2026.02 | 87.97 | |
| BaselineModel=Llama 2-13B2026.02 | 87.85 | |
| KIVISinkModel=Llama 2-13B2026.02 | 87.83 | |
| InnerQBaseModel=Llama 2-13B2026.02 | 87.47 | |
| InnerQSmallModel=Llama 2-13B2026.02 | 84.55 | |
| InnerQBaseModel=Llama 2-7B2026.02 | 83.84 | |
| BaselineModel=Llama 2-7B2026.02 | 83.51 | |
| InnerQHybridModel=Llama 2-7B2026.02 | 83.32 | |
| KIVIModel=Llama 2-7B2026.02 | 82.77 | |
| KIVISinkModel=Llama 2-7B2026.02 | 82.27 | |
| BaselineModel=Llama 3.2-3B2026.02 | 69.39 | |
| KIVISinkModel=Llama 3.2-3B2026.02 | 67.53 | |
| InnerQBaseModel=Llama 3.2-3B2026.02 | 67.48 | |
| KIVIModel=Llama 3.2-3B2026.02 | 67.29 | |
| InnerQSmallModel=Llama 3.2-3B2026.02 | 65.41 | |
| InnerQHybridModel=Llama 3.2-3B2026.02 | 65.37 | |
| KIVISinkModel=Llama 3.2-1B2026.02 | 64.75 | |
| BaselineModel=Llama 3.2-1B2026.02 | 64.03 | |
| KIVIModel=Llama 3.2-1B2026.02 | 63.92 | |
| InnerQBaseModel=Llama 3.2-1B2026.02 | 62.92 | |
| InnerQSmallModel=Llama 3.2-1B2026.02 | 59.2 | |
| InnerQHybridModel=Llama 3.2-1B2026.02 | 59.06 | |
| InnerQSmallModel=Llama 2-7B2026.02 | 39.51 |