Abstractive Summarization Faithfulness on XSum
22.71R-LCD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CDModel=Llama-3-70B2026.06 | 22.71 | 69.99 | 54.73 | |
| ADModel=Llama-3-70B2026.06 | 22.49 | 69.91 | 60.57 | |
| DoLa (low)Model=Llama-3-70B2026.06 | 22.46 | 69.8 | 61.11 | |
| DoLa (high)Model=Llama-3-70B2026.06 | 22.43 | 69.93 | 59.99 | |
| GreedyModel=Llama-3-70B2026.06 | 22.41 | 69.77 | 61.32 | |
| DCOModel=Llama-3-70B2026.06 | 22.4 | 69.74 | 60.87 | |
| DeCoRe (static)Model=Llama-3-70B2026.06 | 21.94 | 69.35 | 64.88 | |
| ITIModel=Llama-3-70B2026.06 | 21.64 | 69.46 | 61.33 | |
| CADModel=Llama-3-70B2026.06 | 21.45 | 69.28 | 65.61 | |
| DCOModel=Llama-3-8B2026.06 | 20.39 | 67.86 | 56.7 | |
| DoLa (high)Model=Llama-3-8B2026.06 | 19.92 | 67.34 | 48.49 | |
| GreedyModel=Llama-3-8B2026.06 | 19.9 | 67.23 | 47.61 | |
| DeCoRe (static)Model=Llama-3-8B2026.06 | 19.87 | 67.83 | 64.07 | |
| DoLa (low)Model=Llama-3-8B2026.06 | 19.82 | 67.19 | 47.21 | |
| ADModel=Llama-3-8B2026.06 | 19.79 | 67.31 | 48.49 | |
| CADModel=Llama-3-8B2026.06 | 18.82 | 67.2 | 67.16 | |
| ITIModel=Llama-3-8B2026.06 | 13.25 | 59.96 | 34.35 |