Skip to content
arXiv cs.LG · Papers

Procedural Fairness Failures in RLHF from Preference Averaging

arXiv:2608.10126v1 Announce Type: new Abstract: Reinforcement Learning from Human Feedback (RLHF) aggregates heterogeneous preferences into a single reward model, assuming preference homogeneity. When preferences are heterogeneous, this aggregation induces a procedural fairness failure where majority preference groups