Roborev maintains a persistent queue of AI-assisted code reviews and the work that produces them. These terms distinguish user-visible reviews from the jobs and panel activity behind them.

Logical review: A standalone review or a panel synthesis parent that appears as one review in user-facing history and metrics. Avoid: Review job when referring to the user-visible result

Review job: One queued agent execution that can produce a review result or contribute to a panel. Avoid: Logical review, experiment

Review panel: A named group of reviewers that examines one target and produces one synthesized logical review. Avoid: Batch, multi-arm review

Panel member: One reviewer within a review panel. Avoid: Experiment arm, sub-review

Synthesis: The panel step that combines member results into the panel's logical review. Avoid: Summary job, evaluator

Review unit: The complete subject of one experiment assignment: either a standalone review or an entire review panel. Avoid: Review job when referring to a panel

Experiment definition: A versioned description of an experiment's eligible workflows, allocation, and experimental review configuration. Avoid: Feature flag, experiment run

Experiment assignment: The immutable placement of one review unit in an experiment arm. Avoid: Experiment definition, eligibility

Default arm: The experiment arm that uses the normal resolved review configuration. Avoid: Control configuration

Experimental arm: The experiment arm that applies the experiment's review-configuration overlay. Avoid: Variant, treatment configuration

Branch subject: A repository-scoped source branch used as the stable identity for experiment assignment and review-session reuse. Avoid: Commit, pull request