Two papers accepted at EWRL 2025: “One Does Not Simply Estimate State: Comparing Model-based and Model-free Reinforcement Learning on the Partially Observable MordorHike Benchmark” introduces a new benchmark for rigorous state estimation testing in POMDPs, and “Mighty: A Comprehensive Tool for studying Generalization, Meta-RL and AutoRL” presents an open-source library unifying contextual generalization, meta-learning, and automated hyperparameter optimization. Looking forward to discussing these papers in Tübingen!