Examine the weakest link
Identify where a conclusion goes beyond direct observation. A valid parser result can feed an unsupported event interpretation; a strong benchmark can still miss the case population. Request evidence that addresses that gap. Explain whether the issue affects wording, method, data, or the conclusion itself.
Protect review integrity
Declare relevant relationships and respect access restrictions on unpublished material. Review approval applies to the submitted version rather than every later change. An editorial reading, citation check, and independent replication are different activities; describe the work performed without giving it a broader status.
Related resources
Canonical reference: https://syntheticforensics.org/reviewers/
Report a technical correction