experiment-plan

To create an experiment plan, you provide a refined research proposal to generate a claim-driven roadmap with run orders, budgets, and risk analyses. This process maps core contributions to concrete experiments, yielding execution-ready outputs for paper submission.

A claim-driven experiment plan is a structured roadmap that directly maps research contributions to evaluation protocols and ablation plans. It enforces explicit success and failure criteria, ensuring every experiment defends a specific claim from the proposal.

Defining run order and compute budgets involves structuring deterministic experiment sequences and allocating data resources. The plan establishes a concise execution roadmap with explicit milestones, ensuring experiments are sequenced logically within resource limits.

Yes, this works for typical data science projects by generating ablation plans alongside the core experiment plan. It structures the evaluation protocol to isolate variables, directly supporting validation requirements for data science research.

You need an execution tracker file to monitor progress and milestones. The workflow generates EXPERIMENT_PLAN.md and EXPERIMENT_TRACKER.md files, providing structured blocks and deterministic run sequencing to document success and failure criteria.

Yes, the experiment planning approach includes risk analysis across the full workflow. It converts your proposal into a detailed plan that addresses potential failures through explicit success criteria and structured risk mitigation blocks.