Classification on Construction Domain (test)
83.92Macro F1Active Learning + Explanation
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Active Learning + ExplanationAL Rounds=3, Num. Manual Labels=68, Accum. Manual Labels=1362, No. Pseudo Labels=2633, Training Samples=3995, Annotation Saved (%)=65.912026.05 | 83.92 | 89.8 | |
| Fine-tunedAL Rounds=-, Num. Manual Labels=3995, Accum. Manual Labels=3995, No. Pseudo Labels=0, Training Samples=3995, Annotation Saved (%)=02026.05 | 83.75 | 90.6 | |
| Active Learning + ExplanationAL Rounds=2, Num. Manual Labels=88, Accum. Manual Labels=1294, No. Pseudo Labels=1701, Training Samples=2995, Annotation Saved (%)=56.792026.05 | 81.6 | 88.4 | |
| Active Learning + ExplanationAL Rounds=0, Num. Manual Labels=1008, Accum. Manual Labels=1008, No. Pseudo Labels=492, Training Samples=1500, Annotation Saved (%)=32.802026.05 | 80.77 | 88.6 | |
| Active Learning + ExplanationAL Rounds=1, Num. Manual Labels=198, Accum. Manual Labels=1206, No. Pseudo Labels=794, Training Samples=2000, Annotation Saved (%)=39.702026.05 | 79.77 | 86.8 | |
| Active LearningAL Rounds=0, Num. Manual Labels=1008, Accum. Manual Labels=1008, No. Pseudo Labels=492, Training Samples=1500, Annotation Saved (%)=32.802026.05 | 76.8 | 84.8 | |
| Zero-shot + ExplanationAL Rounds=-, Num. Manual Labels=0, Accum. Manual Labels=0, No. Pseudo Labels=0, Training Samples=0, Annotation Saved (%)=02026.05 | 43.07 | 52 | |
| Zero-shotAL Rounds=-, Num. Manual Labels=0, Accum. Manual Labels=0, No. Pseudo Labels=0, Training Samples=0, Annotation Saved (%)=02026.05 | 36.05 | 39.8 |