Learning Mixtures of Plackett-Luce Models for Multi-Objective Alignment

arXiv:2608.25200v1 Announce Type: cross Abstract: We consider the problem of learning a mixture of k Plackett-Luce models given multi-way ranking responses from annotators that may represent heterogeneous underlying preferences. This problem has many applications in AI alignment and preference optimization. Prior work has studied mixtures of Bradley-Terry models from pairwise comparisons. However, uncovering mixture models is theoretically unidentifiable when k exceeds m/2, where m is the length of a ranking. We propose an efficient implementation to address this limitation, which involves first augmenting the rankings to a larger size by generating new responses from a base language model, followed by a gradient-based estimation to reduce inference cost in the input embedding space. Based on this procedure, we then design an expectation-maximization algorithm with these two steps to fit a mixture of Plackett-Luce models, called MoPLEx. Extensive experiments are conducted to verify this approach. First, we show that the gradient-based approximation estimates true probabilities with less than 5% error on models with up to 34 billion parameters. Second, we show that MoPLEx improves clustering and ranking accuracy by an average of 43.7% and 15.2% over baselines using single ranking and mixtures of Bradley-Terry models, on preference optimization datasets. These results demonstrate the effectiveness of MoPLEx for tackling multi-way rankings from heterogeneous preferences through measuring alignment between gradients.

Related

• Three Models of RLHF Annotation: Extension, Evidence, and Authority
• Multi-Objective Exploration and Preference Optimization via Mutual Information
• MATO: Multi-objective Personalized Alignment with Test-time Optimization for Large Language Models

Source: arXiv cs.CL | 2026-08-27