Docs
Manual Evals
Manual Evals
Launch reviewer runs from datasets and score rows with pass or fail plus rubric criteria.
Manual Evals
Manual evals are the current evaluation workflow in Captar v1. They let teams review dataset rows with pass or fail verdicts, notes, and weighted rubric scores.
Current workflow
- Create a manual eval from an existing dataset.
- Add reviewer instructions and weighted criteria.
- Start a run.
- Review rows, score criteria, and record verdicts.
Manual review only in v1
Automated evaluators and live evaluation pipelines are intentionally out of scope right now. The shipped workflow is manual review backed by datasets.