Reviews: Efficient state-space modularization for planning: theory, behavioral and neural signatures

The paper is very ambitious and develops a computational model of how the state space can be carved up (aggregated?) for planning. This model is applied to some intriguing data on human and rodent spatial navigation and seems to nicely pull together disparate threads from the literature. Unfortunately, the exposition was sufficiently abstract, so that following the thread from the model to the results (simulations) was challenging, leaving unclear exactly how the model was explaining the behaviour and making evaluation difficult. It is not entirely clear how the different ideas introduced in the paper (modularity, centrality, description length) fit together into a single model of behaviour and the brain. From the text, it was not clear to me how the simulations and predictions for the different behavioural tasks were generated.