Flexibility is one of the hallmarks of human problem-solving. In everyday life, people adapt tochanges in common tasks with little to no additional training. Much of the existing work on flexibil-ity in human problem-solving has focused on how people adapt to tasks in new domains by drawingon solutions from previously learned domains. In real-world tasks, however, humans must generalizeacross a wide range of within-domain variation. In this work we argue that representational abstrac-tion plays an important role in such within-domain generalization. We then explore the nature ofthis representational abstraction in realistically complex tasks like video games by demonstratinghow the same model-based planning framework produces distinct generalization behaviors under dif-ferent classes of task representation. Finally, we compare the behavior of agents with these task rep-resentations to humans in a series of novel grid-based video game tasks. Our results provideevidence for the claim that within-domain flexibility in humans derives from task representationscomposed of propositional rules written in terms of objects and relational categories.