Skip to main content
The Review Copilot is a chat pane on every review page; open it with the Copilot button or ⌘K. It already knows the trace, the rubric, the findings the trace belongs to, its user journey, and the memory saved from earlier reviews, so its help is grounded in the same standard the reviewer is grading against. A review with the Copilot pane open, answering how the trace scores against each item of the rubric

​
What to ask it

Type a question in the pane, or open Copilot Actions next to any section of the trace to point it at one part: the input, the prompt, a single workflow step, or the output. The Copilot Actions menu of a trace's Output section
  • Explain. Ask how a dimension should be applied, or why an example counts as good or bad, so a new reviewer grades it the way an expert would.
  • Draft a rationale. Have it write a one-sentence rationale for a grade or a correction.
  • Draft an edit. Ask for an improved version of the model’s output, then insert it into the review as the corrected response.
  • Propose grades. Ask it to grade the trace against the rubric; it fills in the grades.
  • Check the rubric itself. Ask it to flag gaps, ambiguities, or missing dimensions.

​
Next steps

Judges & Datasets

Calibrate a judge against this review data