CEFR level classification on EVP (test)
83A1 ScoreFT+KB
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| FT+KBBase Model=GPT-4.1 mini, Training Set=EVP2025.10 | 83 | 77 | 74 | 79 | 74 | 81 | 78 | 78 | |
| FT+KBBase Model=GPT-4.1 mini, Training Set=Mixture2025.10 | 83 | 81 | 78 | 83 | 78 | 81 | 81 | 81 | |
| FTBase Model=GPT-4.1 mini, Training Set=Mixture2025.10 | 81 | 76 | 73 | 75 | 61 | 73 | 73 | 73 | |
| FTBase Model=GPT-4.1 mini, Training Set=EVP2025.10 | 79 | 68 | 64 | 69 | 43 | 68 | 65 | 66 | |
| ME6 Cont.Base Model=BERT, Training Set=EVP2025.10 | 77 | 61 | 54 | 53 | 51 | 59 | 59 | 57 | |
| ME6 Cont.Base Model=BERT, Training Set=Mixture2025.10 | 76 | 65 | 59 | 51 | 54 | 59 | 61 | 59 | |
| FT+KBBase Model=GPT-4.1 mini, Training Set=SemCor-CEFR2025.10 | 75 | 72 | 76 | 81 | 75 | 77 | 76 | 76 | |
| FTBase Model=GPT-4.1 mini, Training Set=SemCor-CEFR2025.10 | 72 | 67 | 68 | 71 | 44 | 66 | 67 | 67 | |
| Zero-ShotBase Model=GPT-5, Training Set=—2025.10 | 68 | 44 | 40 | 53 | 29 | 21 | 42 | 45 | |
| 18-ShotBase Model=GPT-5, Training Set=EVP2025.10 | 67 | 45 | 43 | 56 | 38 | 40 | 48 | 49 | |
| 6-ShotBase Model=GPT-5, Training Set=EVP2025.10 | 66 | 44 | 44 | 57 | 40 | 32 | 47 | 49 | |
| ME6 Cont.Base Model=BERT, Training Set=SemCor-CEFR2025.10 | 61 | 51 | 50 | 42 | 46 | 57 | 51 | 50 |