In-Context Learning May Not Elicit Trustworthy Reasoning: A-Not-B Errors in Pretrained Language Models.

Pengrui Han, Peiyang Song, Haofei Yu, Jiaxuan You

Browse the full EMNLP paper archive.