Sentiment Analysis on SST-2 GLUE
94.9F1 ScoreKEN
Evaluation results
| Method | Links | |
|---|---|---|
| KENBackbone=DeBERTa, Trainable params=99M, Reset params (%)=28.352024.02 | 94.9 | |
| KENBackbone=DeBERTa, Trainable params=107M, Reset params (%)=22.842024.02 | 94.8 | |
| KENBackbone=DeBERTa, Trainable params=92M, Reset params (%)=33.862024.02 | 94.6 | |
| KENBackbone=Ernie, Trainable params=57M2024.02 | 94.1 | |
| KENBackbone=Ernie, Trainable params=69M, Reset params (%)=37.052024.02 | 94.1 | |
| KENBackbone=Bert, Trainable params=80M, Reset params (%)=26.552024.02 | 93.8 | |
| KENBackbone=Ernie, Trainable params=75M, Reset params (%)=31.82024.02 | 93.8 | |
| KENTrainable params=80M2024.02 | 93.8 | |
| KENBackbone=Bert, Trainable params=69M, Reset params (%)=37.052024.02 | 93.7 | |
| KENBackbone=Bert, Trainable params=75M, Reset params (%)=31.82024.02 | 93.7 | |
| KENBackbone=Ernie, Trainable params=80M, Reset params (%)=26.552024.02 | 93.4 | |
| Bert baseTrainable params=109M2024.02 | 93.37 | |
| HybridTrainable params=94M2024.02 | 93.23 | |
| KENBackbone=BLOOM1B7, Trainable params=664M, Reset params (%)=61.462024.02 | 92.9 | |
| KENTrainable params=63M2024.02 | 92.9 | |
| KENBackbone=DeBERTa, Trainable params=84M2024.02 | 92.8 | |
| KENBackbone=Bert, Trainable params=57M2024.02 | 92.8 | |
| KENBackbone=BLOOM560M, Trainable params=411M, Reset params (%)=26.342024.02 | 92.4 | |
| KENBackbone=BLOOM560M, Trainable params=429M, Reset params (%)=23.262024.02 | 92.3 | |
| HybridNTTrainable params=94M2024.02 | 92.2 | |
| KENBackbone=BLOOM560M, Trainable params=420M, Reset params (%)=24.82024.02 | 92.1 | |
| KENBackbone=BLOOM560k, Trainable params=404M2024.02 | 92 | |
| KENBackbone=DistilBERT, Trainable params=51M, Reset params (%)=23.452024.02 | 92 | |
| HybridTrainable params=66M2024.02 | 91.97 | |
| KENBackbone=DistilBERT, Trainable params=44M, Reset params (%)=34.392024.02 | 91.9 | |
| KENBackbone=DistilBERT, Trainable params=47M, Reset params (%)=28.922024.02 | 91.9 | |
| KENBackbone=BLOOM1B7, Trainable params=531M2024.02 | 90.9 | |
| KENBackbone=BLOOM1B7, Trainable params=531M, Reset params (%)=69.172024.02 | 90.9 | |
| Gordon et al. (2020)Trainable params=66M2024.02 | 90.8 | |
| HybridNTTrainable params=66M2024.02 | 90.71 | |
| Sajjad et al. (2020)Trainable params=66M2024.02 | 90.3 | |
| KENBackbone=Electra, Trainable params=14M2024.02 | 90.1 | |
| KENBackbone=Electra, Trainable params=14M, Reset params (%)=55.942024.02 | 90.1 | |
| KENBackbone=DistilBERT, Trainable params=40M2024.02 | 89.2 | |
| KENBackbone=Electra, Trainable params=12M, Reset params (%)=64.752024.02 | 85 | |
| FLOPBackbone=Bert, Trainable params=66M2024.02 | 83.2 | |
| FLOPBackbone=Ernie, Trainable params=67M2024.02 | 83.2 | |
| FlopTrainable params=66M2024.02 | 83.2 | |
| FLOPBackbone=DistilBERT, Trainable params=45M2024.02 | 82.4 | |
| FLOPBackbone=DeBERTa, Trainable params=88M2024.02 | 82.3 | |
| FLOPBackbone=BLOOM560k, Trainable params=408M2024.02 | 81.8 | |
| FLOPBackbone=Electra, Trainable params=28M2024.02 | 81.1 | |
| FLOPBackbone=BLOOM1B7, Trainable params=1.1B2024.02 | 80.7 | |
| KENBackbone=BLOOM1B7, Trainable params=442M, Reset params (%)=74.312024.02 | 80.4 | |
| KENBackbone=Electra, Trainable params=8.9M, Reset params (%)=75.562024.02 | 79.9 |