Uncertainty-guided selective generation on CoQA (dev)
96.52AUARC (Token Entropy)UA-CLM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| UA-CLMModel=Gemma-2B2024.12 | 96.52 | 96.68 | 96.71 | 96.72 | |
| UA-CLMModel=Llama-2-13B2024.12 | 96.45 | 97 | 97.84 | 97.92 | |
| UA-CLMModel=Llama-2-7B2024.12 | 96.03 | 96.57 | 96.99 | 97.16 | |
| CLMModel=Llama-2-13B2024.12 | 95.79 | 95.84 | 96.59 | 96.61 | |
| ULTModel=Llama-2-13B2024.12 | 95.1 | 95.34 | 95.92 | 96.12 | |
| CLMModel=Gemma-2B2024.12 | 94.68 | 94.54 | 96.19 | 96.55 | |
| CLMModel=Llama-2-7B2024.12 | 94.35 | 94.44 | 95.08 | 95.3 | |
| Pre-trainedModel=Gemma-2B2024.12 | 92.71 | 93.39 | 92.35 | 94.52 | |
| ULTModel=Llama-2-7B2024.12 | 92.12 | 92.19 | 93.15 | 93.26 | |
| ULTModel=Gemma-2B2024.12 | 91.72 | 91.52 | 90.93 | 91.68 | |
| Pre-trainedModel=Llama-2-13B2024.12 | 86.72 | 89.4 | 91.37 | 92.09 | |
| CTModel=Llama-2-7B2024.12 | 86.03 | 87.8 | 90.29 | 90.75 | |
| CTModel=Llama-2-13B2024.12 | 84.97 | 86.47 | 89.44 | 92 | |
| Pre-trainedModel=Llama-2-7B2024.12 | 83.61 | 86.06 | 93.48 | 94.11 |