If a handful of submissions failed at once — say, during a transient hiccup — you don't have to retry them one at a time. This article shows you the bulk retry.
When an assessment has failed submissions, a Retry failed (N) button appears in the submissions table toolbar, where N is the total number of failed submissions across the whole assessment.
Marking.ai re-queues every eligible failure (transient AI hiccups, infrastructure glitches, stuck jobs), retrying up to 5 at a time so it doesn't overwhelm the AI provider. You'll see those submissions flip from Failed back to Processing, then Marked as they succeed.
Submissions that need a re-upload or an assessment fix first are skipped on purpose — the dialog reports them separately as needing your attention, so you can resolve those individually from the row tooltip or the submission's detail page.
If you re-mark a submission that was already marked and the re-mark fails, the submission keeps its previous marks and shows a Re-mark failed chip instead of Failed. Retry failed doesn't pick these up. To try again, open the row's ⋯ menu and choose Re-mark. See What re-marking costs.
The Activity screen has a Failed tab that lists what's failed across your workspace, so you can see everything in one place. Retry is per-row here (via each row's actions menu) — there's no multi-select bulk retry on Activity. To bulk-retry a whole assessment's failures in one go, use the Retry failed (N) toolbar button on that assessment's submissions table (above).
You can ask Marko to "retry the failed submissions in [assessment]". Marko confirms the scope before running the retry.
Each retry is treated like a fresh marking job: