Question Answering on PopQA (AUROC, AUARC, ECE)
0.908AUROCSGPU
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SGPULLM Architecture=Llama-3.1-8B2025.12 | 0.908 | 0.902 | 0.42 | |
| UMPIRELLM Architecture=Llama-3.1-8B2025.12 | 0.871 | 0.57 | 0.05 | |
| Cos EigenscoreLLM Architecture=Llama-3.1-8B2025.12 | 0.853 | 0.846 | 0.132 | |
| Cov EigenscoreLLM Architecture=Llama-3.1-8B2025.12 | 0.784 | 0.753 | 0.114 | |
| SELLM Architecture=Llama-3.1-8B2025.12 | 0.739 | 0.618 | 0.13 | |
| DSELLM Architecture=Llama-3.1-8B2025.12 | 0.734 | 0.623 | 0.159 | |
| ManualMaternLLM Architecture=Llama-3.1-8B2025.12 | 0.705 | 0.626 | 0.656 | |
| RegularLLM Architecture=Llama-3.1-8B2025.12 | 0.678 | 0.646 | 0.334 | |
| ManualHeatLLM Architecture=Llama-3.1-8B2025.12 | 0.664 | 0.646 | 0.455 |