Question Answering on NQ (EM @5 and @10)
58.31EM Accuracy (@5)DCCD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DCCDModel=Phi-3-medium (14B)2026.07 | 58.31 | 59.56 | |
| CADModel=Phi-3-medium (14B)2026.07 | 58.06 | 59.78 | |
| ADACADModel=Phi-3-medium (14B)2026.07 | 57.95 | 59.72 | |
| fullModel=Phi-3-medium (14B)2026.07 | 57.89 | 59.56 | |
| DVDModel=Phi-3-medium (14B)2026.07 | 57.56 | 58.81 | |
| DCCDModel=Qwen3.5-9B2026.07 | 55.68 | 58.42 | |
| CADModel=Qwen3.5-9B2026.07 | 55.21 | 57.56 | |
| COCOAModel=Phi-3-medium (14B)2026.07 | 55.1 | 54.79 | |
| ADACADModel=Qwen3.5-9B2026.07 | 55.04 | 57.7 | |
| fullModel=Qwen3.5-9B2026.07 | 54.99 | 57.23 | |
| DVDModel=Qwen3.5-9B2026.07 | 54.32 | 55.57 | |
| DCCDModel=Qwen3.5-2B2026.07 | 53.1 | 53.24 | |
| fullModel=Qwen3.5-2B2026.07 | 51.75 | 52.44 | |
| CADModel=Qwen3.5-2B2026.07 | 51.3 | 52.22 | |
| ADACADModel=Qwen3.5-2B2026.07 | 51.05 | 52.05 | |
| COCOAModel=Qwen3.5-9B2026.07 | 50.83 | 51.61 | |
| DVDModel=Qwen3.5-2B2026.07 | 50.25 | 50.3 | |
| COCOAModel=Qwen3.5-2B2026.07 | 45.4 | 45.29 | |
| zero-shotModel=Phi-3-medium (14B)2026.07 | 39.11 | — | |
| zero-shotModel=Qwen3.5-9B2026.07 | 29.56 | — | |
| zero-shotModel=Qwen3.5-2B2026.07 | 15.62 | — |