Next-token prediction on OpenWebText (val)
18.8Perplexity
Hermite activation
Updated 1mo ago
Evaluation Results
| Method | Links | ||
|---|---|---|---|
2025.02 | 18.8 | 2.9337 | |
2025.02 | 18.93 | 2.941 | |
2025.02 | 18.94 | 2.9411 | |
2025.02 | 19.25 | 2.9574 | |
2025.02 | 19.64 | 2.9775 |
Hermite activation
| Method | Links | ||
|---|---|---|---|
2025.02 | 18.8 | 2.9337 | |
2025.02 | 18.93 | 2.941 | |
2025.02 | 18.94 | 2.9411 | |
2025.02 | 19.25 | 2.9574 | |
2025.02 | 19.64 | 2.9775 |