DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing

Benchmark Model Rank Results
natural-language-inference-on-qnliDeBERTaV3large#7Accuracy: 96%
natural-language-inference-on-rteDeBERTaV3large#5Accuracy: 92.7%