Question Answering on HotpotQA (train test)
53.19BLEUNo alteration
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| No alterationProtection=None2026.01 | 53.19 | 55.02 | 28.16 | 54.4 | 63.39 | |
| BAEProtection=Surface-level perturbation2026.01 | 50.99 | 55.07 | 30.45 | 53.12 | 62.02 | |
| Seq2SickProtection=Surface-level perturbation2026.01 | 49.31 | 55.6 | 30.42 | 53 | 61.44 | |
| Disclaimer InjectionProtection=Disclaimer Injection2026.01 | 21.26 | 21.51 | 10.98 | 21.46 | 38.41 |