Delving into the Reversal Curse: How Far Can Large Language Models Generalize?

Neural Information Processing Systems 

A prime example is the recently debated "reversal curse", which surfaces when models, having been trained on the fact "A is B", struggle to generalize this knowledge to infer that "B is A ".

Similar Docs  Excel Report  more

TitleSimilaritySource
None found