Towards human-like questioning: Knowledge base question generation with bias-corrected reinforcement learning from human feedback

Runhao Zhao, Jiuyang Tang, Weixin Zeng, Yunxiao Guo, Xiang Zhao

Abstract: Highlights•Pioneering reinforcement learning from human feedback (RLHF) to enhance knowledge base question generation.•Proposing bias-corrected RLHF to reduce sycophancy and boost question accuracy.•Introducing feedback mechanisms to enhance generated question quality.•Extensive experiments demonstrate our method’s superior performance.