Multi-class generator attribution on SemEval Task 13 Subtask B (test)
39.553Macro F1codegemma-7b
Evaluation results
| Method | Links | |
|---|---|---|
| codegemma-7bModel Size=7B, Training Data Strategy=full dataset with preserved class distribution2026.04 | 39.553 | |
| codegemma-2bModel Size=2B, Training Data Strategy=full dataset with preserved class distribution2026.04 | 36.475 | |
| codegemma-7b_balancedModel Size=7B, Training Data Strategy=undersampled balanced2026.04 | 35.714 | |
| Qwen2.5-Coder-14BModel Size=14B2026.04 | 34.415 | |
| Qwen2.5-Coder-3BModel Size=3B2026.04 | 31.463 | |
| codegemma-2b_balancedModel Size=2B, Training Data Strategy=undersampled balanced2026.04 | 31.188 | |
| CodeBERT baseline2026.04 | 22.858 | |
| random baselineContext=11 classes2026.04 | 9.091 |