Definition
Telling an AI not just whether it scored well, but what went wrong, so it can fix it.
In optimize-anything, structured diagnostic feedback returned by an evaluator alongside the scalar score (failing test IDs, stack traces, profiler output), used as a text-optimization analog of a gradient.