# Response and annotation review

Task and guideline version:
Source evidence:
Candidate output:

For each criterion use: meets / partly meets / fails / not applicable, with evidence.
Factual support:
Instruction following:
Completeness:
Uncertainty handling:
Clarity:

Annotation schema:
Original label / span / box:
Corrected label / span / box:
Evidence for correction:
Ambiguity or disagreement:
Escalation needed:
Preferred response and reason:
Corrected response:
Reviewer / date: