Free-form Text Generation on CoQA
94.61AccuracyUA-CLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| UA-CLMBackbone=Llama-2-13B2024.12 | 94.61 | 0.9118 | 0.0084 | |
| CLMBackbone=Llama-2-13B2024.12 | 94.06 | 0.9106 | 0.0323 | |
| UA-CLMBackbone=Llama-2-7B2024.12 | 92.64 | 0.8882 | 0.0094 | |
| CLMBackbone=Llama-2-7B2024.12 | 92.53 | 0.8886 | 0.0343 | |
| ULTBackbone=Llama-2-13B2024.12 | 92.5 | 0.8771 | 0.0595 | |
| ULTBackbone=Llama-2-7B2024.12 | 89.5 | 0.8409 | 0.0588 | |
| CTBackbone=Llama-2-13B2024.12 | 87.25 | 0.8022 | 0.0992 | |
| Pre-trainedBackbone=Llama-2-13B2024.12 | 85.5 | 0.7832 | 0.0559 | |
| Pre-trainedBackbone=Llama-2-7B2024.12 | 83.5 | 0.7449 | 0.0561 | |
| CTBackbone=Llama-2-7B2024.12 | 81.25 | 0.7437 | 0.041 | |
| DeLaskModel=LLaMA3-8B, Decoding=DeLask2026.05 | 68.05 | — | — | |
| ITIModel=LLaMA3-8B, Decoding=ITI2026.05 | 67.71 | — | — | |
| RegularModel=LLaMA3-8B, Decoding=Regular2026.05 | 67.4 | — | — | |
| UADModel=LLaMA3-8B, Decoding=UAD2026.05 | 64.67 | — | — | |
| DeLaskModel=LLaMA2-7B, Decoding=DeLask2026.05 | 63.76 | — | — | |
| ADModel=LLaMA3-8B, Decoding=AD2026.05 | 63.4 | — | — | |
| DOLAModel=LLaMA3-8B, Decoding=DOLA2026.05 | 62.91 | — | — | |
| RegularModel=LLaMA2-7B, Decoding=Regular2026.05 | 62.55 | — | — | |
| ITIModel=LLaMA2-7B, Decoding=ITI2026.05 | 62.48 | — | — | |
| ADModel=LLaMA2-7B, Decoding=AD2026.05 | 62.38 | — | — | |
| UADModel=LLaMA2-7B, Decoding=UAD2026.05 | 61.65 | — | — | |
| DOLAModel=LLaMA2-7B, Decoding=DOLA2026.05 | 60.76 | — | — |