Entity-Swap Question Answering on NQ-Swap
84.7Exact Match (EM)CAD
Evaluation Results
| Method | Links | |
|---|---|---|
| CADModel=Llama-3-70B2026.06 | 84.7 | |
| DeCoRe (static)Model=Llama-3-70B2026.06 | 79.06 | |
| GreedyModel=Llama-3-70B2026.06 | 76.11 | |
| DoLa (low)Model=Llama-3-70B2026.06 | 75.98 | |
| DCOModel=Llama-3-70B2026.06 | 75.33 | |
| DoLa (high)Model=Llama-3-70B2026.06 | 75.26 | |
| ITIModel=Llama-3-70B2026.06 | 74.76 | |
| CADModel=Llama-3-8B2026.06 | 74.21 | |
| ADModel=Llama-3-70B2026.06 | 74.02 | |
| CDModel=Llama-3-70B2026.06 | 68.37 | |
| DeCoRe (static)Model=Llama-3-8B2026.06 | 64.43 | |
| DCOModel=Llama-3-8B2026.06 | 62.96 | |
| DoLa (high)Model=Llama-3-8B2026.06 | 60.98 | |
| DoLa (low)Model=Llama-3-8B2026.06 | 60.77 | |
| GreedyModel=Llama-3-8B2026.06 | 60.62 | |
| ADModel=Llama-3-8B2026.06 | 60.51 | |
| ITIModel=Llama-3-8B2026.06 | 51.08 |