Text Generation Evaluation Correlation on WebText (test)
0.643Perplexity (PPL)
Bradley-Terry Score (Interesting)
Updated 4d ago
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| 0.643 | 0.524 | -0.143 | 52.4 | 40.5 | 0.81 | ||
| 0.738 | 0.69 | -0.071 | 59.5 | 52.4 | 0.857 | ||
| 0.81 | 0.833 | -0.167 | 73.8 | 59.5 | 0.952 |