Vetted probabilistic earthquake forecasts can contribute to more earthquake-resilient societies. Forecasts underpin seismic hazard assessments and thus determine building and life safety. They also provide scientifically sound information about the time-dependence of earthquake potential before, during and after earthquake sequences. To ensure forecasts are trustworthy and to assess the scientific hypotheses underlying the forecasts, models should be tested both retrospectively and prospectively (i.e., against yet-to-be-collected data). For this purpose, the Collaboratory for the Study of Earthquake Predictability (CSEP) provides tools and methods for testing the consistency and precision of earthquake forecasts. This session welcomes contributions that showcase advances in the science of earthquake forecasting and model testing. These can include: new approaches for identifying precursory activity (e.g. b-value variations, aseismic slip transients); forecasts based on empirical machine-learning or physical stress-transfer algorithms; applications of models to earthquake sequences around the globe; advances in model evaluation techniques; or contributions to software tools for model developers. Presentations may also highlight progress of community efforts, such as the EU H2020 project RISE (Real-time earthquake rIsk reduction for a reSilient Europe, www.rise-eu.org) and other initiatives.