Interaction-Grounded Learning for Contextual Markov Decision Processes with Personalized Feedback

TL;DR

This paper introduces a new algorithm for Interaction-Grounded Learning in sequential decision-making, enabling personalized feedback handling in complex MDPs with provable guarantees and practical effectiveness.

Contribution

It extends IGL to multi-step MDPs with personalized feedback, providing a computationally efficient algorithm with regret guarantees for realistic sequential tasks.

Findings

The proposed algorithm achieves sublinear regret in episodic MDPs.

Effective in learning personalized objectives from multi-turn interactions.

Validated on synthetic and real-world datasets.

Abstract

In this paper, we study Interaction-Grounded Learning (IGL) [Xie et al., 2021], a paradigm designed for realistic scenarios where the learner receives indirect feedback generated by an unknown mechanism, rather than explicit numerical rewards. While prior work on IGL provides efficient algorithms with provable guarantees, those results are confined to single-step settings, restricting their applicability to modern sequential decision-making systems such as multi-turn Large Language Model (LLM) deployments. To bridge this gap, we propose a computationally efficient algorithm that achieves a sublinear regret guarantee for contextual episodic Markov Decision Processes (MDPs) with personalized feedback. Technically, we extend the reward-estimator construction of Zhang et al. [2024a] from the single-step to the multi-step setting, addressing the unique challenges of decoding latent rewards under MDPs. Building on this estimator, we design an Inverse-Gap-Weighting (IGW) algorithm for policy optimization. Finally, we demonstrate the effectiveness of our method in learning personalized objectives from multi-turn interactions through experiments on both a synthetic episodic MDP and a real-world user booking dataset.