1 paper
Shaokai He, Kaiwen Wei, Xinyi Zeng +5
The "reversal curse" refers to the phenomenon where large language models (LLMs) exhibit predominantly unidirectional behavior when processing logically bidirectional relationships…