Question Answering on NQ-open Augmented (full-slice)
85.42Restate-hardNWCAD
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| NWCADModel=Llama-3.1-70B2026.04 | 85.42 | 46 | 63.2 | 68.97 | 68.75 | |
| NWCADModel=Llama-3.1-8B2026.04 | 84.4 | 66 | 64.09 | 51.4 | 54.6 | |
| AdaCADModel=Llama-3.1-70B2026.04 | 80.56 | 30.4 | 61.42 | 64.59 | 62.62 | |
| AdaCADModel=Llama-3.1-8B2026.04 | 80 | 31.4 | 55.39 | 45.87 | 51.2 | |
| With-contextModel=Llama-3.1-8B2026.04 | 77.6 | 36.2 | 51.93 | 43.8 | 50.6 | |
| NWCADModel=Ministral-3-8B2026.04 | 75.8 | 40 | 61.42 | 44.2 | 53.8 | |
| CoCoAModel=Llama-3.1-8B2026.04 | 72.8 | 64.6 | 59.94 | 50.98 | 53 | |
| AdaCADModel=Ministral-3-8B2026.04 | 72 | 33.4 | 58.16 | 46 | 48.4 | |
| With-contextModel=Llama-3.1-70B2026.04 | 69.68 | 36.4 | 43.92 | 49.81 | 51.75 | |
| With-contextModel=Ministral-3-8B2026.04 | 69.2 | 34.2 | 45.4 | 42 | 49.8 | |
| CoCoAModel=Llama-3.1-70B2026.04 | 68.75 | 16.8 | 57.57 | 63.7 | 63.75 | |
| CoCoAModel=Ministral-3-8B2026.04 | 63.2 | 26.4 | 56.97 | 45.4 | 50.6 | |
| CADModel=Llama-3.1-8B2026.04 | 60.4 | 11.4 | 53.12 | 46.8 | 49.4 | |
| BaselineModel=Llama-3.1-8B2026.04 | 58.4 | 60.4 | 2.97 | 0 | 0 | |
| CADModel=Ministral-3-8B2026.04 | 54.4 | 19.8 | 50.74 | 36.6 | 44.4 | |
| CADModel=Llama-3.1-70B2026.04 | 42.82 | 26.6 | 43.03 | 57.2 | 51.12 | |
| BaselineModel=Llama-3.1-70B2026.04 | 41.67 | 44 | 7.12 | 0 | 0.12 | |
| BaselineModel=Ministral-3-8B2026.04 | 32.4 | 36.2 | 5.04 | 0 | 0 |