Summarization on RadNLI (ROUGE-L)
39.6ROUGE-LFull FT
Evaluation results
| Method | Links | |
|---|---|---|
| Full FTBase model=gpt-oss 20B, Trainable Parameters=100.0%2026.04 | 39.6 | |
| MPTBase model=gpt-oss 20B, Trainable Parameters=<0.05%2026.04 | 39.1 | |
| Full FTBase model=Meditron3 8B, Trainable Parameters=100.0%2026.04 | 37.9 | |
| LoRABase model=gpt-oss 20B, Trainable Parameters=~2.50%2026.04 | 37.7 | |
| MPTBase model=Meditron3 8B, Trainable Parameters=<0.05%2026.04 | 37.2 | |
| Full FTBase model=LLaMA 3.1 8B, Trainable Parameters=100.0%2026.04 | 36.4 | |
| LoRABase model=Meditron3 8B, Trainable Parameters=~2.50%2026.04 | 35.8 | |
| MPTBase model=LLaMA 3.1 8B, Trainable Parameters=<0.05%2026.04 | 35.7 | |
| LoRABase model=LLaMA 3.1 8B, Trainable Parameters=~2.50%2026.04 | 34.2 | |
| PTBase model=gpt-oss 20B, Trainable Parameters=<0.05%2026.04 | 34 | |
| PTBase model=Meditron3 8B, Trainable Parameters=<0.05%2026.04 | 31.9 | |
| PTBase model=LLaMA 3.1 8B, Trainable Parameters=<0.05%2026.04 | 30.1 |