rlxp-plan-baseline

To plan a reinforcement learning baseline run, you need to define training and evaluation commands, seeds, expected outputs, and metric extraction paths for your active study contract. This process creates a reproducible baseline candidate file and execution summary before launch.

A study-scoped baseline plan is a concrete run preparation document that defines the first lifecycle-controlled training run. It captures commands, seeds, and metric extraction paths to remove manual guesswork from turning approved study requirements into a reproducible execution summary.

Yes, preparing an RL baseline candidate requires task cards, adapters, contracts, and hardware and budget details. These inputs are necessary to generate the baseline candidate file and command summary without altering existing task code or metrics.

No, baseline run preparation is safety-bounded and strictly keeps setup separate from reward, curriculum, domain randomization, task code, and metric changes. This scoping ensures the baseline plan remains a controlled first run before any autoloop or candidate validation.

The best way to define metric extraction paths is by generating a baseline candidate file through a structured run preparation process. This captures expected outputs and evaluation commands, ensuring metrics are correctly traced before initiating autoloop or run analysis.

You should not use automated baseline planning if your study workflow is not yet approved or if you intend to simultaneously modify reward, curriculum, domain randomization, or task code. It is designed strictly for first lifecycle-controlled runs requiring no code changes.