rlxp-propose-candidates

To choose the next reinforcement-learning experiment, you convert run analysis into bounded candidate proposals for interventions like reward tuning or curriculum learning. This ensures your next hypothesis stays aligned with the active task, study budget, and validation evidence.

The best way to plan safe reward tuning and curriculum learning interventions is to generate validation-ready YAML outputs from contract-approved objectives. This keeps proposals bounded and aligned with the current study budget and allowed intervention classes.

Yes, you need a study contract to propose domain randomization candidates. The process requires contract context, state and ledger inputs, and lifecycle metadata to ensure proposed interventions do not drift out of the approved scope or budget.

Generating validation-ready YAML for RL experiment planning requires contract context, state and ledger inputs, hypothesis framing, and lifecycle metadata. These inputs ensure the proposed candidates are safe for downstream execution within the allowed intervention classes.

Yes, you can use this approach to address instrumentation repair issues in your reinforcement-learning workflow. It converts run analysis into bounded candidate proposals, ensuring any repair interventions remain within the active study contract and budget constraints.

You should avoid automated candidate generation when you lack a contract-approved task scope, budget, or lifecycle metadata. Without these foundational inputs, the generated reinforcement-learning experiment proposals cannot guarantee safe downstream execution or scope compliance.