| paraphrase-identification-on-mrpc-glue-dev | FFN GEGLU | #1 | F1: 92.68 |
| natural-language-inference-on-rte-glue-dev | FFN GLU | #8 | Accuracy: 84.12 |
| semantic-textual-similarity-on-sts-b-glue-dev | FFN SwiGLU | #8 | Pearson/Spearman Correlation: 90.23 |
| natural-language-understanding-on-glue-dev | FFN ReGLU | #10 | GLUE (dev): 84.67 |
| linguistic-acceptability-on-cola-glue-dev | FFN ReGLU | #11 | Matthews Correlation: 56.16 |
| natural-language-inference-on-multinli-dev | FFN Bilinear | #12 | Matched: 86.90 |
| natural-language-inference-on-qnli-glue-dev | FFN SwiGLU | #12 | Accuracy: 92.93 |
| question-answering-on-boolq | FFN Bilinear | #38 | Accuracy: 81.53 |
| question-answering-on-squad11-dev | FFN ReGLU | #55 | F1: 91.18 |
| sentiment-analysis-on-sst-2-binary-classification | FFN ReGLU | #89 | Dev Accuracy: 94.38 |