Uncertainty-guided selective generation on TriviaQA (val)
92AUARC (Token Entropy)UA-CLM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| UA-CLMModel=Llama-2-13B2024.12 | 92 | 92.54 | 91.55 | 92.52 | |
| ULTModel=Llama-2-13B2024.12 | 89.49 | 90.55 | 90.13 | 90.63 | |
| UA-CLMModel=Llama-2-7B2024.12 | 88.79 | 89.27 | 87.8 | 89.34 | |
| CLMModel=Llama-2-13B2024.12 | 87.98 | 88.07 | 87.08 | 88.29 | |
| CLMModel=Llama-2-7B2024.12 | 86.17 | 86.15 | 85.58 | 86.3 | |
| CTModel=Llama-2-7B2024.12 | 85.71 | 88.12 | 87.69 | 88.34 | |
| CTModel=Llama-2-13B2024.12 | 84.6 | 89.14 | 90.07 | 90.19 | |
| Pre-trainedModel=Llama-2-13B2024.12 | 84.41 | 85.74 | 85.84 | 85.87 | |
| ULTModel=Llama-2-7B2024.12 | 82.73 | 84.18 | 84.48 | 85.19 | |
| Pre-trainedModel=Llama-2-7B2024.12 | 80.5 | 82.59 | 82.51 | 83.69 | |
| UA-CLMModel=Gemma-2B2024.12 | 73.73 | 74.36 | 72.58 | 74.53 | |
| CLMModel=Gemma-2B2024.12 | 72.56 | 72.51 | 71.62 | 71.98 | |
| ULTModel=Gemma-2B2024.12 | 72.12 | 74.29 | 72.46 | 74.13 | |
| Pre-trainedModel=Gemma-2B2024.12 | 69.12 | 72.25 | 71.27 | 72.79 |