Skip to main content

​
Review while the run is working

Question Studio shows candidates as individual workers finish, so you can begin reading the first question while the run continues. The run header shows the destination, source scope, question mix, current phase, and overall progress. Pending slots remain visible until they produce a candidate or are cut.

​
Inspect each candidate

Select a question from the left-hand list. On smaller screens, it opens in a modal. Review:
  • question type and topic
  • stem and options
  • marked correct answer
  • rationale
  • cited pages and quoted source evidence
  • duplicate or quality warnings
Use the up and down arrow keys to browse and the space bar to select or clear the active candidate while focus is outside a form control.

​
Understand automated checks

Question Studio applies source, answer-structure, and duplicate checks. Depending on the run path, a candidate may also receive an independent model review. When a card says that only local evidence and structure checks ran, manually verify correctness and distractor quality before publishing. Automated checks reduce obvious failures, while content expertise supplies the final judgment.

​
Edit a candidate

After the run is ready, click the edit control in the inspector to change:
  • the stem
  • answer options
  • the correct option
  • the rationale
Apply or cancel the edit before saving drafts. Curator edits are recorded without replacing the original generation and source provenance.

​
Select what to keep

Use the checkbox on each candidate, or choose All or None above the list. A selected candidate is queued to save. Publication remains a separate module action. Candidates that Question Studio cuts appear in the run details with a reason when one is available. Common reasons include weak source support, ambiguity, duplication, or an unsupported requested question type.

​
Read run details

The run diagnostics summarize operational information such as:
  • drafted versus planned questions
  • kept and cut candidates
  • model calls and failures
  • tokens and estimated spend
  • coverage across planned topics
  • the activity timeline
Cost values can be partial or estimated. Use them for operational monitoring; evaluate quality through source and content review.

​
Save selected drafts

When the run is ready, click Save n drafts to module. Question Studio saves only selected, unsaved candidates with draft status. After saving:
  • the saved candidates are labeled Saved
  • they can no longer be selected again in that run
  • Open module takes you to the destination and highlights the generated drafts
  • publishing remains a separate action in the module

​
Start another run safely

Use New run or Start a new run after you have saved what you want. If unsaved candidates remain, Question Studio asks whether to keep reviewing or discard them.

​
Final publishing checklist

Before publishing from the module, confirm:
  1. The correct answer is unambiguous.
  2. The rationale explains the relationship behind the answer.
  3. Every required fact is supported by the cited source.
  4. Distractors are plausible without becoming alternate correct answers.
  5. The question adds a distinct learning target to the module.