Self-training
Reported on 1 benchmark across 1 task
Note: results are matched by exact model name. Different papers may use the same name for different model variants.
Natural Language Processing1 result
- F1 score· uses extra data92.1best: 96.43 (Hashing + XLNet)