SNLI
The Stanford Natural Language Inference (SNLI) dataset is a benchmark released by Stanford University containing textual entailment, contradiction, and neutral label annotations. It is one of the most widely used reasoning evaluation datasets in NLP, though subsequent work identified shortcut-learning issues within it.
Timeline (last 90 days)
NLI is a fundamental task in natural language processing: given a premise and a hypothesis, determine whether the relationship is entailment, contradiction or neutral, usually based on pretrained models such as BERT, DeBERTa and RoBERTa fine-tuned on the MNLI and SNLI datasets.
Laya achieves an accuracy of 0.86 on the SNLI English natural language inference dataset.
All Facts (2)
NLI is a fundamental task in natural language processing: given a premise and a hypothesis, determine whether the relationship is entailment, contradiction or neutral, usually based on pretrained models such as BERT, DeBERTa and RoBERTa fine-tuned on the MNLI and SNLI datasets.
50%UnverifiedLaya achieves an accuracy of 0.86 on the SNLI English natural language inference dataset.
50%