Testers cannot see inside a pottery-glaze photo model. They raise a similar model whose internals they do know, search that stand-in, then replay the interesting inputs on the original, assuming shared decision boundaries. What approach is that?
Select an answer to reveal the explanation.
Short Explanation
Raising a similar model whose internals testers know, searching that stand-in, then replaying the interesting inputs on the original is black-box adversarial testing that uses an equivalent model and transferability. It is not white-box on the original, not static drift testing, and not back-to-back winner selection.
Full Explanation
The black-box path uses an equivalent model with known internals, then transfers cases to the original. That is not white-box on the original, not static drift testing, and not A/B-style winner selection.