Human Review

Review and resolve low-confidence evaluation tasks.

Human Review is where reviewers decide executions that score below evaluator confidence thresholds. Open Control Center → Evals → Human Review.

Tasks move from Review Pending to Reviewed after a reviewer records a verdict.

Human Review is the queue for tasks awaiting a reviewer verdict.

Use My tasks, All tasks, and Unassigned to switch queues. Filter by team, type, status, or priority.

The queue summary shows Assigned to me, Reviews in view, Reviewed, and Unassigned. Select View unassigned to open unassigned tasks.

Each task shows:

• Execution, Eval type, and Evaluator that flagged it
• Team and Assignee responsible for the task
• Priority, Queued time, Status, and recorded Verdict

Open a task, record a verdict, and move it to Reviewed.

The open task's evaluation summary shows the judge's reasoning and, when the judge gave any, its Evidence: up to three points, each citing something the agent did or failed to do and whether it raised or lowered the score.

Assignments use teams

Human Review assigns tasks through teams, not directly to individuals. A reviewer without a team has no assigned tasks.

Configure teams in Admin.

Each evaluator has its own confidence threshold. For example, one evaluator can route scores below 67%, while another routes scores below 100%.

---

Maintained by Abhishek Paul AI-assisted, human-approved