Tag
Researchers introduce a method to automatically augment commonsense knowledge corpora with negation, creating 2M+ triples that improve LLM negation understanding when used for pre-training.