Multilingual Few-shot Text Classification on Multilingual NLP Benchmark Suite (test)
84.3M-SST-2 AccuracyEFL
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| EFLProtocol=Translate Train + Translate Test, Few-shot K=8, Backbone=XLM-R2021.04 | 84.3 | 81.5 | 90.5 | 83.3 | 88.7 | 73.1 | 67.6 | 81.3 | |
| EFLProtocol=Train + Translate Test, Few-shot K=8, Backbone=XLM-R2021.04 | 84.1 | 81.2 | 90.1 | 83.2 | 84.5 | 72.9 | 68.7 | 80.7 | |
| Standard FTProtocol=Translate Train + Translate Test, Few-shot K=8, Backbone=XLM-R2021.04 | 55.1 | 63.8 | 65.5 | 57 | 89.4 | 69.4 | 69.1 | 67 | |
| EFL wo PTProtocol=Translate Train + Translate Test, Few-shot K=8, Backbone=XLM-R2021.04 | 54.6 | 60.7 | 65.6 | 59.3 | 89.5 | 71.3 | 69.2 | 67.2 | |
| EFL wo PTProtocol=Train + Translate Test, Few-shot K=8, Backbone=XLM-R2021.04 | 52.6 | 54.3 | 57.2 | 59.9 | 76.5 | 67 | 69.2 | 62.4 | |
| Standard FTProtocol=Train + Translate Test, Few-shot K=8, Backbone=XLM-R2021.04 | 52.2 | 54.3 | 55.8 | 56.9 | 77.5 | 67.2 | 69.2 | 61.9 | |
| MajorityBackbone=XLM-R2021.04 | 50.9 | 50 | 50 | 50 | 50 | 66.8 | 69.1 | 55.3 |