RLHF and IIA: Perverse Incentives
Wanqiao Xu · Shi Dong · Xiuyuan Lu · Grace Lam · Zheng Wen · Benjamin Van Roy
2024 Poster
in
Workshop: Models of Human Feedback for AI Alignment
in
Workshop: Models of Human Feedback for AI Alignment
Abstract
Existing algorithms for reinforcement learning from human feedback (RLHF) can incentivize responses at odds with preferences because they are based on models that assume independence of irrelevant alternatives (IIA). The perverse incentives induced by IIA hinder innovations on query formats and learning algorithms.
Video
Chat is not available.
Successful Page Load