Browse documentationReviewer cohorts
Choose who should review the task
Evaluation access and staffing are separate. A task can remain available to the verified fal.ai team or use a saved Standard reviewer cohort defined by demographic, language, role, device preference, and coarse technical-context filters. Creative Engineer and MLOpt access follows the task's internal staff policy rather than pretending those roles belong to the matched Standard cohort.
| Choice | Who can enter | Use it when |
|---|---|---|
| fal.ai team | Verified @fal.ai accounts with an eligible workspace role | The default for broad internal evaluation |
| Saved rater cohort | Standard reviewers matched by one immutable filter snapshot | The task needs a reusable, bounded internal cohort |
| Internal staff access | Creative Engineer and MLOpt users allowed by the task policy | Expert review or task operations are required |
| External assignment | One explicitly invited external account | A collaborator must review one task without workspace access |
Saved cohort filters
A rater cohort is a saved filter, not a hand-picked directory of people. Current filters include workspace role, self-reported country, primary language, age band, hearing support, preferred display and media, plus coarse browser family, operating-system family, approximate network country, viewport bucket, screen bucket, and connection class.
The preview returns only a thresholded aggregate count. Saving materializes an immutable cohort version so later profile or device changes do not rewrite an active task's staffing history. Creating a newer cohort version is the supported way to change the filter.
Profile and technical context
Reviewers enter stable profile fields during onboarding. GenMedia also stores a coarse browser, operating system, locale, timezone, viewport, screen, network-country, and connection-class context under the accepted Terms. Exact IP addresses, exact screen dimensions, browsing history, and arbitrary browser payloads are not part of the staffing contract.
A new Evaluation Session freezes the allowed technical context needed for later quality review. Hearing and vision information remains separately gated because it is accessibility data. Session-time playback checks still decide whether the current device can satisfy an audio or video protocol.
Filter and preflight serve different purposes
- Saved cohort
Selects eligible Standard reviewers from stable or coarse profile context before Sessions begin.
- Session preflight
Checks the current browser, media playback, audio channel, and task-specific readiness before counted review.
- Reviewer quality
Uses server-derived reliability and experience evidence; it is not inferred from age, country, or device.
- Staff override
Creative Engineer and MLOpt participation is reported separately and never inflated into the matched-cohort count.
External reviewers
External accounts are available only through explicit assignment. They see the minimum context required to evaluate the candidates and never receive general workspace or result access. Do not assign confidential material to an external reviewer unless that sharing is approved outside GenMedia as well.