Long-Horizon Model-Based Offline Reinforcement Learning Without Explicit Conservatism

TL;DR

This paper introduces NEUBAY, a Bayesian approach to offline RL that avoids explicit conservatism, enabling long-horizon rollouts and achieving state-of-the-art results on benchmarks.

Contribution

It presents a novel Bayesian method for offline RL that handles epistemic uncertainty without explicit conservatism, allowing effective long-horizon planning.

Findings

NEUBAY outperforms conservative algorithms on 7 datasets.

Long-horizon rollouts are crucial when conservatism is removed.

Bayesian approach excels on low-quality datasets.

Abstract

Popular offline reinforcement learning (RL) methods rely on explicit conservatism, penalizing out-of-dataset actions or restricting rollout horizons. We question the universality of this principle and revisit a complementary Bayesian perspective for test-time adaptation. By modeling a posterior over world models and training a history-dependent agent to maximize expected return, the Bayesian approach directly addresses epistemic uncertainty without explicit conservatism. We first illustrate in a bandit setting that Bayesianism excels on low-quality datasets where conservatism fails. Scaling to realistic tasks, we find that long-horizon rollouts are essential to control value overestimation once conservatism is removed. We introduce design choices that enable learning from long-horizon rollouts while mitigating compounding model errors, yielding our algorithm, NEUBAY, grounded in the neutral Bayesian principle. On D4RL and NeoRL benchmarks, NEUBAY is competitive with leading conservative algorithms, achieving new state-of-the-art on 7 datasets with rollout horizons of several hundred steps. Finally, we characterize datasets by quality and coverage to identify when NEUBAY is preferable to conservative methods.