Your first document review
Start with one document task and a small set of cases whose answers you can check.
1. Request access
Use the early-access request on the homepage. Access is reviewed manually. After you receive an invitation, create your account, confirm your email, and create a workspace.
2. Define the task and evidence
Create a project with a clear task and success rule. Use a small, permitted document sample, add cases and expected answers, and keep development and final cases separate. Do not start with confidential material you are not authorized to share.
3. Request the required review
Dataset approval is a staff step. Submit the dataset for review; resolve requested changes before freezing an approved version. This is not an instant self-service training service.
4. Compare the same cases
For the first review, import baseline and candidate answers as JSONL against the same frozen cases. Imports record supplied answers; they do not prove which model ran. Inspect failures, regressions and source references, and record any required human judgments.
5. Save and inspect the result
Save an eligible reviewed comparison as a report, then open the report and inspect its versions and unresolved cases. A saved report is evidence of this review, not a guarantee of model quality or deployment safety.
Questions or support: contact@frontierintelligencelab.com or contact us.