Question Answering on HotpotQA (full-slice)
53.8HotpotQA Full-Slice Distance (d)NWCAD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NWCADModel=Llama-3.1-8B2026.04 | 53.8 | 62.2 | |
| NWCADModel=Llama-3.1-70B2026.04 | 53.28 | 60.38 | |
| NWCADModel=Ministral-3-8B2026.04 | 52.6 | 60.4 | |
| AdaCADModel=Ministral-3-8B2026.04 | 50.4 | 58.2 | |
| CoCoAModel=Ministral-3-8B2026.04 | 50 | 60.2 | |
| AdaCADModel=Llama-3.1-8B2026.04 | 48.4 | 60.8 | |
| CoCoAModel=Llama-3.1-8B2026.04 | 46.6 | 57.6 | |
| CoCoAModel=Llama-3.1-70B2026.04 | 46.38 | 59.5 | |
| AdaCADModel=Llama-3.1-70B2026.04 | 44.5 | 61.25 | |
| CADModel=Ministral-3-8B2026.04 | 43.4 | 53.2 | |
| With-contextModel=Llama-3.1-70B2026.04 | 40.38 | 47.62 | |
| CADModel=Llama-3.1-8B2026.04 | 39.6 | 53.6 | |
| With-contextModel=Llama-3.1-8B2026.04 | 39.2 | 43.4 | |
| With-contextModel=Ministral-3-8B2026.04 | 36.4 | 46.8 | |
| CADModel=Llama-3.1-70B2026.04 | 29.38 | 47.62 | |
| BaselineModel=Llama-3.1-8B2026.04 | 14.8 | 14.8 | |
| BaselineModel=Ministral-3-8B2026.04 | 12 | 12 | |
| BaselineModel=Llama-3.1-70B2026.04 | 11.75 | 11.75 |