Definition Generation on Multi-word tokens famous people, places, entities, sayings and concepts
99.2CorrectnessOriginal tokenization
Evaluation results
| Method | Links | |
|---|---|---|
| Original tokenizationModel=Llama3-8B-i2025.05 | 99.2 | |
| Original tokenizationModel=Llama3.1-8B-i2025.05 | 97.8 | |
| Original tokenizationModel=OLMo2-7B-i2025.05 | 94 | |
| Original tokenizationModel=Mistral-7B2025.05 | 93.4 | |
| Original tokenizationModel=Llama3.1-8B2025.05 | 91.1 | |
| Original tokenizationModel=Llama3.2-3B-i2025.05 | 90.9 | |
| Original tokenizationModel=Avg.2025.05 | 90 | |
| Original tokenizationModel=Llama3-8B2025.05 | 88.1 | |
| Token DistillationModel=Llama3-8B-i2025.05 | 86.3 | |
| Token Distillation + αNTPModel=Llama3-8B-i2025.05 | 85.7 | |
| Token Distillation + NTPModel=Llama3-8B-i2025.05 | 85.3 | |
| Token DistillationModel=OLMo2-7B-i2025.05 | 84.3 | |
| Token Distillation + αNTPModel=OLMo2-7B-i2025.05 | 83.7 | |
| Token DistillationModel=Llama3.1-8B-i2025.05 | 81.3 | |
| Token Distillation + NTPModel=OLMo2-7B-i2025.05 | 80.9 | |
| Token Distillation + αNTPModel=Llama3.1-8B-i2025.05 | 80.1 | |
| Token Distillation + NTPModel=Llama3.1-8B-i2025.05 | 79.7 | |
| NTPModel=Llama3.1-8B-i2025.05 | 71.6 | |
| Token Distillation + αNTPModel=Llama3-8B2025.05 | 71 | |
| NTPModel=Llama3-8B-i2025.05 | 70.6 | |
| NTPModel=OLMo2-7B-i2025.05 | 70 | |
| Token DistillationModel=Llama3-8B2025.05 | 68 | |
| Token Distillation + NTPModel=Llama3-8B2025.05 | 68 | |
| Token Distillation + αNTPModel=Avg.2025.05 | 66.6 | |
| Token Distillation + αNTPModel=Llama3.2-3B-i2025.05 | 66.4 | |
| Token Distillation + NTPModel=Avg.2025.05 | 65.5 | |
| Original tokenizationModel=Llama3.2-3B2025.05 | 65.2 | |
| Token DistillationModel=Avg.2025.05 | 64.6 | |
| Token Distillation + NTPModel=Llama3.2-3B-i2025.05 | 63.4 | |
| Token Distillation + αNTPModel=Llama3.1-8B2025.05 | 62 | |
| Token DistillationModel=Llama3.1-8B2025.05 | 61.8 | |
| Token Distillation + NTPModel=Llama3.1-8B2025.05 | 60.2 | |
| Token DistillationModel=Llama3.2-3B-i2025.05 | 57.5 | |
| NTPModel=Llama3-8B2025.05 | 54.3 | |
| NTPModel=Avg.2025.05 | 54.1 | |
| NTPModel=Llama3.2-3B-i2025.05 | 53.9 | |
| NTPModel=Llama3.1-8B2025.05 | 49.7 | |
| Token Distillation + NTPModel=Mistral-7B2025.05 | 48.3 | |
| Token Distillation + αNTPModel=Mistral-7B2025.05 | 46.9 | |
| Token DistillationModel=Mistral-7B2025.05 | 46.3 | |
| Token Distillation + NTPModel=Llama3.2-3B2025.05 | 38.2 | |
| Token Distillation + αNTPModel=Llama3.2-3B2025.05 | 37.2 | |
| NTPModel=Mistral-7B2025.05 | 34 | |
| Token DistillationModel=Llama3.2-3B2025.05 | 31.6 | |
| NTPModel=Llama3.2-3B2025.05 | 29 | |
| ZeTTModel=Llama3-8B-i2025.05 | 28.4 | |
| ZeTTModel=Llama3-8B2025.05 | 25.4 | |
| ZeTTModel=Mistral-7B2025.05 | 15.3 | |
| MeanModel=Llama3.1-8B-i2025.05 | 11.3 | |
| MeanModel=Llama3-8B-i2025.05 | 9.1 | |
| MeanModel=Llama3.1-8B2025.05 | 8 | |
| MeanModel=Llama3.2-3B-i2025.05 | 7.4 | |
| MeanModel=OLMo2-7B-i2025.05 | 7.4 | |
| MeanModel=Avg.2025.05 | 6.7 | |
| MeanModel=Llama3-8B2025.05 | 4.2 | |
| MeanModel=Llama3.2-3B2025.05 | 4.2 | |
| MeanModel=Mistral-7B2025.05 | 2.2 | |
| RandomModel=Llama3.2-3B-i2025.05 | 0.2 | |
| RandomModel=Mistral-7B2025.05 | 0 | |
| RandomModel=Llama3-8B2025.05 | 0 | |
| RandomModel=Llama3-8B-i2025.05 | 0 | |
| RandomModel=Llama3.1-8B2025.05 | 0 | |
| RandomModel=Llama3.1-8B-i2025.05 | 0 | |
| RandomModel=Llama3.2-3B2025.05 | 0 | |
| RandomModel=OLMo2-7B-i2025.05 | 0 | |
| RandomModel=Avg.2025.05 | 0 |