Continuous learning changes weights and biases and can raise or lower robustness. Changes should be assessed with a predefined set of criteria, comparing the outcomes for the new version of the network against the previous version, and a further criterion should be set on the observed difference so that any degradation can be judged acceptable or not.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.