Natural Language Processing on SuperGLUE 1k samples, excl. ReCoRD (dev)
65.84Macro Avg ScoreMulti-CLS BERT
Evaluation Results
| Method | Links | |
|---|---|---|
| Multi-CLS BERTBackbone=BERT Large, Model Size=350.9M, K=5, lambda=0.52022.10 | 65.84 | |
| Multi-CLS BERTBackbone=BERT Large, Model Size=350.9M, K=5, lambda=0.12022.10 | 65.58 | |
| Multi-CLS BERTBackbone=BERT Large, Model Size=350.9M, K=5, lambda=02022.10 | 65.43 | |
| MTLBackbone=BERT Large, Model Size=335.2M2022.10 | 65.21 | |
| Multi-CLS BERTBackbone=BERT Large, Model Size=350.9M, K=5, lambda=12022.10 | 65 | |
| Multi-CLS BERTBackbone=BERT Large, Model Size=338.3M, K=12022.10 | 64.67 | |
| Multi-CLS BERTBackbone=BERT Base, Model Size=118.4M, K=5, lambda=0.52022.10 | 63.78 | |
| Multi-CLS BERTBackbone=BERT Base, Model Size=118.4M, K=5, lambda=02022.10 | 63.71 | |
| Multi-CLS BERTBackbone=BERT Base, Model Size=118.4M, K=5, lambda=0.12022.10 | 63.61 | |
| Multi-CLS BERTBackbone=BERT Base, Model Size=118.4M, K=5, lambda=12022.10 | 63.56 | |
| Multi-CLS BERTBackbone=BERT Base, Model Size=111.3M, K=12022.10 | 63.35 | |
| MTLBackbone=BERT Base, Model Size=109.5M2022.10 | 62.94 | |
| PretrainedBackbone=BERT Base, Model Size=109.5M2022.10 | 61.55 |