Natural Language Inference on MNLI Matched (test)
90.2AccuracyRoBERTa-Large
Evaluation Results
| Method | Links | |
|---|---|---|
| RoBERTa-LargeParameters (approx.)=355M, Fine-tuning=true2026.03 | 90.2 | |
| ALBERT-LargeParameters (approx.)=17M, Fine-tuning=true2026.03 | 86.76 | |
| RoBERTa-BaseParameters (approx.)=125M, Fine-tuning=true2026.03 | 85.98 | |
| BERT-LargeParameters (approx.)=340M, Fine-tuning=true2026.03 | 84.83 | |
| BERT-BaseParameters (approx.)=110M, Fine-tuning=true2026.03 | 80.99 | |
| ALBERT-BaseParameters (approx.)=11M, Fine-tuning=true2026.03 | 79.89 | |
| NGT (Our Model)Parameters (approx.)=110M, Fine-tuning=true2026.03 | 79 | |
| ESIMParameters (approx.)=4.3M, Fine-tuning=true2026.03 | 76.63 | |
| DAMParameters (approx.)=382K, Fine-tuning=true2026.03 | 68.41 |