Not expected but reasonable, if there is coupling between the concepts of malicious code and malicious other activities, through some sort of generalized unders

Not expected but reasonable, if there is coupling between the concepts of malicious code and malicious other activities, through some sort of generalized understanding/information-conceptual-compression in the "knowledge ensemble" One experiment could be to repeat this across models of varying size and see if the bigger models (assuming trained on ~similar dataset) are more capable of conceptual compartmentalization