RA

Human review

Adjudication

Review low-confidence or high-variance model evaluations with prompt context, peer output, rubric metadata, and a structured decision.

Queue

0 items require review.

IDLE

Queue clear

No evaluations currently require manual adjudication.

Current evaluation

Prompt, output, expected answer, rubric, and peer variance.

IDLE

No evaluation selected

Select a queued item to inspect prompt context, output, rubric, and peer variance.

Decision

Persist a structured score override and review note.

Scores are stored as explicit overrides.