Masked Language Modeling on 6 languages Averaged (test)
42.7MRRNoOverlap
Evaluation Results
| Method | Links | |
|---|---|---|
| NoOverlapTokenizer method=NoOverlap, Probing=Average of 5 random seeds2023.05 | 42.7 | |
| UnigramTokenizer method=Unigram, Probing=Average of 5 random seeds2023.05 | 42 | |
| TokMixTokenizer method=TokMix, Probing=Average of 5 random seeds2023.05 | 38.7 | |
| BPETokenizer method=BPE, Probing=Average of 5 random seeds2023.05 | 35.6 |