UBP2: Uncertainty-Balanced Preference Planning for Efficient Preference-based Reinforcement Learning
For RL practitioners, this provides a principled way to improve sample efficiency in preference-based RL without ad hoc exploration heuristics.
UBP2 introduces a model-based preference-based RL method that actively balances exploration and exploitation by jointly reasoning over uncertainties in reward, dynamics, and value functions, achieving sublinear regret guarantees and substantially higher sample efficiency on Meta-World benchmarks compared to prior methods.
Preference-based RL provides an approach to learning reward models from pairwise comparisons of behaviors, bypassing the need for explicit reward design. However, existing methods typically rely on passive data collection and suffer from poor sample efficiency, especially during the early stages of learning. We introduce a model-based approach that actively directs exploration by jointly reasoning over uncertainties in the reward, dynamics, and value functions. Our method, Uncertainty-Balanced Preference Planning (UBP2), uses ensembles of reward, dynamics, and value function models to evaluate candidate trajectories according to a unified score that combines expected reward, terminal value, and epistemic uncertainty. Planning under this objective yields an explicit tradeoff between exploitation and information acquisition without requiring ad hoc exploration heuristics. Under standard regularity assumptions, we establish sublinear regret guarantees for both finite-horizon and infinite-horizon settings. Empirically, experiments on the Meta-World benchmark show UBP2 achieves substantially higher sample efficiency than model-free preference-based methods and non-optimistic model-based baselines.