Text Compression on Estonian Corpus
4.46Compression RatioLlama-3.2-1B
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-3.2-1BTokenizer Strategy=prune+ext (16k, ours)2025.12 | 4.46 | |
| Llama-3.2-3BTokenizer Strategy=prune+ext (16k, ours)2025.12 | 4.46 | |
| Llama-3.2-1BTokenizer Strategy=prune+ext (16k, naive)2025.12 | 4.21 | |
| Llama-3.2-3BTokenizer Strategy=prune+ext (16k, naive)2025.12 | 4.21 | |
| Llama-3.2-1BTokenizer Strategy=default2025.12 | 2.64 | |
| Llama-3.2-3BTokenizer Strategy=default2025.12 | 2.64 |