From Reward-Free Representations to Preferences: Rethinking Offline Preference-Based Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Jun-Jie, Hsu, Chia-Heng, Chen, Kui-Yuan, Hsieh, Ping-Chun
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917551298052096
author Yang, Jun-Jie
Hsu, Chia-Heng
Chen, Kui-Yuan
Hsieh, Ping-Chun
author_facet Yang, Jun-Jie
Hsu, Chia-Heng
Chen, Kui-Yuan
Hsieh, Ping-Chun
contents Preference-based reinforcement learning (PbRL) avoids explicit reward engineering by learning from pairwise human preference feedback. Existing offline PbRL methods typically follow a two-stage pipeline, first learning a reward or preference model from labeled preferences and then performing offline RL on unlabeled data. We revisit offline PbRL through the lens of reward-free representation learning (RFRL) from the zero-shot RL literature, and propose a new training framework that first learns latent successor-measure representations from reward-free offline data, followed by contrastive search and fine-tuning using preference data. Through extensive experiments and ablations, we show that our method achieves superior preference efficiency over offline PbRL baselines. This work is the first to connect RFRL with PbRL, highlighting its potential as a feedback-efficient solution. Our code is publicly available at https://github.com/rl-bandits-lab/FB-PbRL.
format Preprint
id arxiv_https___arxiv_org_abs_2606_01123
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle From Reward-Free Representations to Preferences: Rethinking Offline Preference-Based Reinforcement Learning
Yang, Jun-Jie
Hsu, Chia-Heng
Chen, Kui-Yuan
Hsieh, Ping-Chun
Machine Learning
Preference-based reinforcement learning (PbRL) avoids explicit reward engineering by learning from pairwise human preference feedback. Existing offline PbRL methods typically follow a two-stage pipeline, first learning a reward or preference model from labeled preferences and then performing offline RL on unlabeled data. We revisit offline PbRL through the lens of reward-free representation learning (RFRL) from the zero-shot RL literature, and propose a new training framework that first learns latent successor-measure representations from reward-free offline data, followed by contrastive search and fine-tuning using preference data. Through extensive experiments and ablations, we show that our method achieves superior preference efficiency over offline PbRL baselines. This work is the first to connect RFRL with PbRL, highlighting its potential as a feedback-efficient solution. Our code is publicly available at https://github.com/rl-bandits-lab/FB-PbRL.
title From Reward-Free Representations to Preferences: Rethinking Offline Preference-Based Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2606.01123