Learning safe action models with conditional effects may require an exponential number of samples, but Conditional-SAM offers a solution.
