You have two candidate models: A is strong on five of s...

When comparing models with different performance profiles across trust axes, the choice depends on the context of the deployment. Simply looking at aggregate radar area can be misleading if the underlying data is manipulated to fit a narrative. Zeroing out specific axes distorts the aggregate performance for risks that might not be relevant to the current deployment, while removing axes entirely hides the fact that a constraint was ever considered. Instead, evaluate the models based on the specific requirements of the use case to ensure the chosen model aligns with the necessary safety and performance standards.