The public contents places supervised learning first among the approaches, which reflects its role as the reference case in the framework. The model is given inputs and the corresponding target, and learns a mapping that can be applied to new inputs. The quality of the labels sets the ceiling on what can be learned, and the split between training, validation and test data sets the honesty of the result. The framework's data and pipeline clauses carry the consequences: labels are a data preparation activity, and the test set must stay out of the training and tuning loop. The clause text is not in the public sample and is not restated here.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.