Authorship Boundary Detection on RoFT chatgpt (leave-one-out cross-domain)
0.52Accuracy (Speeches)Phi1.5 (GB)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Phi1.5 (GB)Pred.=Perpl., Context=sent., Classification Method=Gradient Boosting2024.10 | 0.52 | 0.241 | 0.457 | 0.561 | |
| GigaCheck (DN-DAB-DETR)Pred.=Text, Context=global2024.10 | 0.5 | 0.328 | 0.545 | 0.636 | |
| Phi1.5 (LR)Pred.=Perpl., Context=sent., Classification Method=Logistic Regression2024.10 | 0.412 | 0.211 | 0.452 | 0.515 | |
| RoBERTaPred.=Text, Context=global2024.10 | 0.363 | 0.148 | 0.38 | 0.361 | |
| RoBERTa SEPPred.=Text, Context=global2024.10 | 0.314 | 0.131 | 0.381 | 0.286 | |
| TLE TS BinaryPred.=TLE, Context=20 tkn2024.10 | 0.147 | 0.163 | 0.171 | 0.111 | |
| PHD TS multiPred.=PHD, Context=100 tkn2024.10 | 0.131 | 0.195 | 0.172 | 0.176 |