Part-of-Speech Tagging on DialectBench (test)
58.35aln POS AccuracyXLM-R + VAÇAÍ-Bowl
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| XLM-R + VAÇAÍ-BowlBackbone=XLM-R, Source selection method=TOPPing2026.05 | 58.35 | 34.88 | 5.47 | 54.02 | 55.16 | 69.97 | 60.81 | 55.17 | 60.21 | 29.23 | |
| XLM-R + VAÇAÍ-BowlBackbone=XLM-R, Source selection method=LangRank2026.05 | 57.97 | 32.94 | 6.24 | 55.97 | 51.71 | 63.54 | 60.35 | 53.83 | 50.46 | 24.97 | |
| XLM-R + AlignmentBackbone=XLM-R, Source selection method=TOPPing2026.05 | 57.81 | 34.08 | 4.95 | 55.36 | 53.22 | 68.7 | 61.34 | 54.54 | 58.35 | 29.22 | |
| XLM-R + AlignmentBackbone=XLM-R, Source selection method=LangRank2026.05 | 57.17 | 33.37 | 6.17 | 55.25 | 50.48 | 61.15 | 61.18 | 53.11 | 47.4 | 24.64 | |
| XLM-RBackbone=XLM-R, Source selection method=TOPPing2026.05 | 56.91 | 33.11 | 7.33 | 53.87 | 53.73 | 67.98 | 61.95 | 53.89 | 60.16 | 31.07 | |
| XLM-RBackbone=XLM-R, Source selection method=LangRank2026.05 | 56.34 | 31.54 | 5.01 | 54.66 | 50.87 | 62.57 | 60.2 | 53.83 | 48.74 | 25.28 | |
| mBERT + VAÇAÍ-BowlBackbone=mBERT, Source Selection Method=TOPPing2026.05 | 50.5 | 33.53 | 10.89 | 49.34 | 48.21 | 69.3 | 66.41 | 43.99 | 61.25 | 25.93 | |
| mBERT + AlignmentBackbone=mBERT, Source Selection Method=LangRank2026.05 | 47.79 | 32.46 | 11.69 | 41.48 | 46.47 | 60.56 | 68.25 | 37.38 | 57.16 | 22.14 | |
| mBERTBackbone=mBERT, Source Selection Method=LangRank2026.05 | 45.76 | 32.07 | 12.08 | 41.63 | 45.63 | 62.61 | 66.54 | 39.93 | 58.64 | 22.61 | |
| mBERT + AlignmentBackbone=mBERT, Source Selection Method=TOPPing2026.05 | 45.57 | 33.34 | 9.51 | 49.4 | 48.53 | 68.7 | 63.93 | 44.06 | 61.22 | 28.26 | |
| mBERT + VAÇAÍ-BowlBackbone=mBERT, Source Selection Method=LangRank2026.05 | 45.51 | 32.17 | 14.54 | 42.71 | 45.55 | 60.48 | 68.36 | 36.38 | 56.89 | 23.81 | |
| mBERTBackbone=mBERT, Source Selection Method=TOPPing2026.05 | 43.8 | 33.37 | 11.19 | 48.68 | 47.71 | 68.32 | 65.48 | 42.03 | 59.26 | 26.56 |