# Refined PAC-Bayes Bounds for Offline Bandits

arXiv: https://arxiv.org/abs/2502.11953 · HTML: https://mlsanity.com/paper/2502.11953/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 26 references that are in the map.
- First posted on arXiv 17 Feb 2025.

## Builds on (most cited first, 7 of 7)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Trust Region Policy Optimization](https://mlsanity.com/paper/1502.05477/) (2015)
- [Counterfactual Risk Minimization: Learning from Logged Bandit Feedback](https://mlsanity.com/paper/1502.02362/) (2015)
- [More PAC-Bayes bounds: From bounded losses, to losses with general tail behaviors, to anytime validity](https://mlsanity.com/paper/2306.12214/) (2023)
- [Information Complexity and Generalization Bounds](https://mlsanity.com/paper/2105.01747/) (2021)
- [An Information-Theoretic Approach to Generalization Theory](https://mlsanity.com/paper/2408.13275/) (2024)
- [PAC-Bayes Bounds for Bandit Problems: A Survey and Experimental Comparison](https://mlsanity.com/paper/2211.16110/) (2022)

## Built on by (most cited first, 1 of 1)

- [PAC-Bayesian Reward-Certified Outcome Weighted Learning](https://mlsanity.com/paper/2604.01946/) (2026)

## Abstract (arXiv)

In this paper, we present refined probabilistic bounds on empirical reward estimates for off-policy learning in bandit problems. We build on the PAC-Bayesian bounds from Seldin et al. (2010) and improve on their results using a new parameter optimization approach introduced by Rodr'iguez et al. (2024). This technique is based on a discretization of the space of possible events to optimize the "in probability" parameter. We provide two parameter-free PAC-Bayes bounds, one based on Hoeffding-Azuma's inequality and the other based on Bernstein's inequality. We prove that our bounds are almost optimal as they recover the same rate as would be obtained by setting the "in probability" parameter after the realization of the data.