Withdrawal and retraction

Drafts and unpublished reviews

An unpublished draft can be archived or deleted from your dashboard so it is no longer an active working copy. It is not in the public sitemap and is not listed on papers or frameworks.

Published work: retract, do not pretend it never existed

Retraction is a formal withdrawal from the public scientific record on this site. The page stays up with a retraction notice. That is intentional. A published, reviewed claim that later comes down still happened; hiding the URL would let someone pretend the timestamp and review never existed.

Retraction is irreversible in the product. Use it when you are withdrawing the claim, not when you want to keep iterating. Iteration is edit + re-review, which writes a new version on top of the last one.

What the audit trail keeps

• The submission text that was reviewed, as a snapshot
• The review report, scores, and specialist reports
• Prediction timestamps once locked_at is set
• Version history after a re-review

Visibility toggles can hide scores or reports from public visitors. They do not delete the stored review. Account closure requests are handled by support; review snapshots tied to a public timestamp are retained for integrity even if the listing is retracted.

Why a finished review is not opted out retroactively

Once models have read a submission and written a report, that inference already happened under the provider terms in force at the time. TOE-Share does not use submissions to train its own models, and it does not grant providers a training license. What we cannot do is rewind a completed API call or erase a timestamp that was offered as priority protection.

If you need the work off the public index, retract or keep it unpublished. If you need the scientific priority, leave the timestamp in place. See priority defense and the terms.