Given only the page's definitions, do two compilers put narrowed curriculum in base loss, incidence, or both?

Is the hierarchy a single-agent ranking, or a claim about which agent's ranking binds when they conflict?

Has any result crossed the domain line and forced a different instrument, or is the remainder still a correspondence of names?

Does listing an instrument that the same page has already shown being gamed count as engineering the vector?

What observation in unplanted models would show that the incidence-shifting share of passing rises with eval-context recognition?

What would count as the measured representation being unable to relocate, rather than merely being inspected?