Factuality Error Localization on FRANK
56.7Accuracy (OutE)Bart-Large
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Bart-LargeProtocol=Fine-tuned on FRANK, Backbone=Bart-Large2024.07 | 56.7 | 36.9 | 14.8 | 34 | 21.4 | 0 | 40 | 29.1 | |
| FineSurEBackbone=GPT-42024.07 | 50.2 | 63.7 | 41.9 | 38.1 | 44.6 | 19.4 | 37.8 | 42.2 | |
| Random GuessingDescription=Randomly selecting from the seven categories (1/7 probability)2024.07 | 14.3 | 14.3 | 14.3 | 14.3 | 14.3 | 14.3 | 14.3 | 14.3 |