Hallucination Detection on OpenDialKG Eval (test)
76.2Macro F1Our approach
Evaluation results
| Method | Links | |
|---|---|---|
| Our approachDetection Type=T5-base Finetuned over Synthetic Data, Base Model=T5-base, Latency=0.20 sec2024.07 | 76.2 | |
| HaluEvalDetection Type=T5-base Finetuned over Synthetic Data, Base Model=T5-base, Latency=0.20 sec2024.07 | 70.2 | |
| SCALE_XLDetection Type=Zero-shot Detection, Latency=0.22 sec2024.07 | 68.7 | |
| FADEDetection Type=T5-base Finetuned over Synthetic Data, Base Model=T5-base, Latency=0.20 sec2024.07 | 62.5 | |
| G-EvalDetection Type=Zero-shot Detection, Latency=2.79 sec2024.07 | 60.8 | |
| SelfCheckGPT (NLI)Detection Type=Zero-shot Detection, Latency=0.93 sec2024.07 | 57.9 | |
| SelfCheckGPT (QA)Detection Type=Zero-shot Detection, Latency=60.59 sec2024.07 | 53.6 |