Language Modeling on FineWeb
48.78PerplexityBitToken
Evaluation Results
| Method | Links | |
|---|---|---|
| BitTokenNumber encoding strategy=BitToken2025.10 | 48.78 | |
| Single DigitNumber encoding strategy=Single Digit2025.10 | 64.23 | |
| SubwordNumber encoding strategy=Subword2025.10 | 74.83 | |
| FoNENumber encoding strategy=FoNE2025.10 | 79.78 | |
| xValNumber encoding strategy=xVal2025.10 | 102.57 |