Machine Translation on FLORES EN-ET v1 (test)
83.5COMETLlama-3.2-3B
Evaluation results
| Method | Links | |
|---|---|---|
| Llama-3.2-3BTraining Status=Continued pretraining, Tokenizer Strategy=default, Training Tokens=7.2B2025.12 | 83.5 | |
| Llama-3.2-3BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, ours), Training Tokens=5.3B2025.12 | 82.2 | |
| Llama-3.2-3BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, naive), Training Tokens=5.4B2025.12 | 81.8 | |
| Llama-3.2-1BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, ours), Training Tokens=5.3B2025.12 | 79.6 | |
| Llama-3.2-1BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, naive), Training Tokens=5.4B2025.12 | 79 | |
| Llama-3.2-1BTraining Status=Continued pretraining, Tokenizer Strategy=default, Training Tokens=7.2B2025.12 | 78.8 | |
| Llama-3.2-3BTraining Status=No training, Tokenizer Strategy=default, Training Tokens=02025.12 | 63.5 | |
| Llama-3.2-1BTraining Status=No training, Tokenizer Strategy=default, Training Tokens=02025.12 | 43.2 |