1 paper · 1 filter
Zhouhao Sun, Xiao Ding, Li Du +5
Despite significant progress, recent studies indicate that current large language models (LLMs) may still capture dataset biases and utilize them during inference, leading to the p…