Domain Shift Extractive Question Answering on AdversarialQA (test)
0.075ECESBA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SBABackbone=LLaMA-2-7B2026.02 | 0.075 | 0.782 | |
| Deep-EnsBackbone=LLaMA-2-7B2026.02 | 0.088 | 0.756 | |
| Temp. Scal.Backbone=LLaMA-2-7B2026.02 | 0.097 | — | |
| Gauss+ProjBackbone=LLaMA-2-7B2026.02 | 0.102 | 0.729 | |
| Laplace-LoRABackbone=LLaMA-2-7B2026.02 | 0.108 | 0.718 | |
| LoRABackbone=LLaMA-2-7B2026.02 | 0.142 | 0.658 |