Guardado en:
Detalles Bibliográficos
Autores principales: Lee, Junghyun, Hong, Minju, Jun, Kwang-Sung, Yun, Chulhee, Yun, Se-Young
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2602.23116
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915836735782912
author Lee, Junghyun
Hong, Minju
Jun, Kwang-Sung
Yun, Chulhee
Yun, Se-Young
author_facet Lee, Junghyun
Hong, Minju
Jun, Kwang-Sung
Yun, Chulhee
Yun, Se-Young
contents We consider the problem of contextual online RLHF with general preferences, where the goal is to identify the Nash Equilibrium. We adopt the Generalized Bilinear Preference Model (GBPM) to capture potentially intransitive preferences via low-rank, skew-symmetric matrices. We investigate general preference learning with any strongly convex regularizer and regularization strength $η^{-1}$, generalizing beyond prior work limited to reverse KL-regularization. Central to our analysis is proving that the dual gap of the greedy policy is bounded by the square of the estimation error, a result derived solely from strong convexity and the skew-symmetry of GBPM. Building on this insight and a feature diversity assumption, we establish two regret bounds via two simple algorithms: (1) Greedy Sampling achieves polylogarithmic, $e^{\mathcal{O}(η)}$-free regret $\tilde{\mathcal{O}}(ηd^4 (\log T)^2)$. (2) Explore-Then-Commit achieves $\mathrm{poly}(d)$-free regret $\tilde{\mathcal{O}}(\sqrt{ηr T})$ by exploiting the low-rank structure; this is the first statistically efficient guarantee for online RLHF in high-dimensions.
format Preprint
id arxiv_https___arxiv_org_abs_2602_23116
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Regularized Online RLHF with Generalized Bilinear Preferences
Lee, Junghyun
Hong, Minju
Jun, Kwang-Sung
Yun, Chulhee
Yun, Se-Young
Machine Learning
Computer Science and Game Theory
We consider the problem of contextual online RLHF with general preferences, where the goal is to identify the Nash Equilibrium. We adopt the Generalized Bilinear Preference Model (GBPM) to capture potentially intransitive preferences via low-rank, skew-symmetric matrices. We investigate general preference learning with any strongly convex regularizer and regularization strength $η^{-1}$, generalizing beyond prior work limited to reverse KL-regularization. Central to our analysis is proving that the dual gap of the greedy policy is bounded by the square of the estimation error, a result derived solely from strong convexity and the skew-symmetry of GBPM. Building on this insight and a feature diversity assumption, we establish two regret bounds via two simple algorithms: (1) Greedy Sampling achieves polylogarithmic, $e^{\mathcal{O}(η)}$-free regret $\tilde{\mathcal{O}}(ηd^4 (\log T)^2)$. (2) Explore-Then-Commit achieves $\mathrm{poly}(d)$-free regret $\tilde{\mathcal{O}}(\sqrt{ηr T})$ by exploiting the low-rank structure; this is the first statistically efficient guarantee for online RLHF in high-dimensions.
title Regularized Online RLHF with Generalized Bilinear Preferences
topic Machine Learning
Computer Science and Game Theory
url https://arxiv.org/abs/2602.23116