Natural Language Inference on SNLI synthetic contrast (val)
90.4AccuracyNaive Finetuning
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Naive FinetuningModel Strategy=Naive Finetuning2025.12 | 90.4 | 81 | |
| Dynamic Balanced SamplingModel Strategy=Dynamic Balanced2025.12 | 90.3 | 81 | |
| Baseline (Pre-trained)Model Strategy=Baseline (Pre-trained)2025.12 | 81.6 | 63.5 |