Grammatical Construct Attempt Detection on W&I-EGP
90.46F1 (General)GPT-4.1
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4.1Statement Type=Lexical, Corrections=Manual2026.03 | 90.46 | 91.03 | 81.61 | |
| GPT-4.1Statement Type=All, Corrections=Manual2026.03 | 86.16 | 85.93 | 73.61 | |
| GPT-4.1Statement Type=Non-lexical, Corrections=Manual2026.03 | 81.87 | 80.83 | 65.62 | |
| QwenStatement Type=Lexical, Corrections=Manual, Model Version=2.5 32B2026.03 | 81.18 | 80.98 | 73.63 | |
| QwenStatement Type=All, Corrections=Manual, Model Version=2.5 32B2026.03 | 76.97 | 75.58 | 64.25 | |
| RBStatement Type=Lexical, Corrections=Manual2026.03 | 76.62 | 76.26 | 65.09 | |
| RBStatement Type=All, Corrections=Manual2026.03 | 73.68 | 72.39 | 56.57 | |
| QwenStatement Type=Non-lexical, Corrections=Manual, Model Version=2.5 32B2026.03 | 72.77 | 70.19 | 54.87 | |
| RBStatement Type=Non-lexical, Corrections=Manual2026.03 | 70.75 | 68.53 | 48.05 |