Summarization on Medical Random subset
25.04R-LCSVOCABADAPT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VOCABADAPTBackbone=Qwen2.5-7B, Best Ckpts.=3500, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 25.04 | 1.09 | 1.1 | 75.72 | |
| VOCABADAPTBackbone=Llama-3.1-8B, Best Ckpts.=3500, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 24.98 | 1.05 | 1.06 | 75.98 | |
| VOCABADAPTBestsetting=zero-shot2026.05 | 24.98 | — | — | 75.98 | |
| CPTOnly (No Vocab Adapt)Backbone=Llama-3.1-8B, Best Ckpts.=7500, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 24.89 | 1.16 | 1.16 | 75.55 | |
| CPTOnly (No Vocab Adapt)Backbone=Qwen2.5-7B, Best Ckpts.=8000, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 24.73 | 1.19 | 1.23 | 75.32 | |
| CPTOnly (No Vocab Adapt)Backbone=Llama-3.1-8B, Best Ckpts.=7500, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 23.92 | 1.2 | 1.34 | 75.43 | |
| VOCABADAPTBackbone=Llama-3.1-8B, Best Ckpts.=3500, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 23.65 | 1.06 | 1.12 | 75.58 | |
| VOCABADAPTBackbone=Qwen2.5-7B, Best Ckpts.=3500, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 23.12 | 1.1 | 1.14 | 75.21 | |
| Llama-3.1-8B-BASEBest Ckpts.=-, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 23.03 | 1.16 | 1.16 | 70.66 | |
| CPTOnly (No Vocab Adapt)Backbone=Qwen2.5-7B, Best Ckpts.=8000, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 22.41 | 1.24 | 1.36 | 74.44 | |
| GPT-5-minisetting=zero-shot2026.05 | 21.8 | — | — | 73.89 | |
| Llama-3.1-8B-BASEBest Ckpts.=-, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 21.33 | 1.2 | 1.34 | 70.41 | |
| Qwen2.5-7B-BASEBest Ckpts.=-, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 15.51 | 1.19 | 1.23 | 43.35 | |
| Qwen2.5-7B-BASEBest Ckpts.=-, Evaluation Protocol=In-context learning, Demonstration=One exemplar2026.05 | 12.23 | 1.24 | 1.36 | 35.22 |