Hallucination Detection on FEVER
59.54AccuracySemantic Entropy
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Semantic EntropyLatency (s/sample)=0.314 ± 0.003, GPU Usage (%)=3.835 ± 0.0122026.05 | 59.54 | 64.77 | |
| Token-level EntropyLatency (s/sample)=0.189 ± 0.002, GPU Usage (%)=3.765 ± 0.0102026.05 | 48.98 | 46.33 | |
| HalluciNotLatency (s/sample)=1.460 ± 0.020, GPU Usage (%)=24.75 ± 0.1502026.05 | 33.48 | 94.11 |