Causal-Guided Active Learning for Debiasing Large Language Models

Li Du, Zhouhao Sun, Xiao Ding +5

Although achieving promising performance, recent analyses show that current generative large language models (LLMs) may still capture dataset biases and utilize them for generation…