Skip to content
← Back to changelog

Changelog

Evaluation review and dataset sign-offs

Production guidance · Console v0.2 · Updated

Released 12 October 2026

Console release v0.2.2 updates Lab Review in the v0.2 interface. Evaluation results can enter the review queue for Judge disagreement or Judge failed, with the request and recorded candidate answer available for human review. Reviewers can approve, correct or reject an answer, add optional rubric scores and notes, and save approved answers or corrections to a Lab dataset. Judge scores stay hidden while the item is pending. The queue supports assignment, changing a decision and reopening an item. Reopening preserves the decision history and removes the example previously saved by that review. If another reviewer changes the item first, the Console asks you to refresh before saving. Read Review evaluation results in Lab Review for the steps and limits.

Sources and scope

Reviewed against a verified production release. The release record and source-code evidence are retained internally. Check the Console version and workspace availability in your environment.