Summarization on Medical domain (100 random samples)
3.3CoherenceLlama-VOCABADAPT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama-VOCABADAPTBase Model=Llama2026.05 | 3.3 | 3.82 | 3.84 | |
| Qwen-VOCABADAPTBase Model=Qwen2026.05 | 3.24 | 3.92 | 3.94 | |
| Llama-CPTOnlyBase Model=Llama, Vocabulary Adaptation=No Vocab Adapt2026.05 | 2.7 | 3.3 | 3.48 | |
| Qwen-CPTOnlyBase Model=Qwen, Vocabulary Adaptation=No Vocab Adapt2026.05 | 2.56 | 3.34 | 3.76 |