<ul><li>Multi-Objective Reinforcement Learning (MORL) is a prominent field of research for balancing trade-offs in real-world sequential decision-making tasks.</li><li>Existing MORL literature lacks focus on generalization across diverse environments, which is crucial in multi-objective contexts.</li><li>This paper introduces a benchmark for evaluating generalization in MORL and evaluates state-of-the-art algorithms, revealing limited capabilities in generalization.</li><li>The study highlights the need for multi-objective specifications and addresses algorithmic complexities to improve generalization in MORL.</li></ul>

On Generalization Across Environments In Multi-Objective Reinforcement Learning

Discover more