1 paper · 1 filter
Peiying Yu, Guoxin Chen, Jingjing Wang
Despite the remarkable capabilities of large language models (LLMs) in various reasoning tasks, they still struggle with table reasoning tasks, particularly in maintaining consiste…