Proposes an adaptive quantile schedule in Bayesian risk MDPs for online RL that starts robust and gradually encourages exploration, supported by asymptotic normality characterization and sublinear Bayesian regret bounds.
Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, volume 108 of Proceedings of Machine Learning Research, 1431--1441
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Evolving Robustness--Exploration Trade-off in Online Reinforcement Learning via Quantile Bayesian Risk MDPs
Proposes an adaptive quantile schedule in Bayesian risk MDPs for online RL that starts robust and gradually encourages exploration, supported by asymptotic normality characterization and sublinear Bayesian regret bounds.