ablation-planner

An ablation study systematically removes model components or varies hyperparameters to validate claims for a research paper. This Skill generates the structured plan from a reviewer perspective, detailing experiments, datasets, and workflows to ensure paper-ready results.

You plan hyperparameter variations and component removals by listing them as input context. The system then parses this input to generate an executable workflow and a structured Ablation Plan detailing the exact run order and feasibility checks.

Structuring ablation experiments from a reviewer perspective involves creating a step-by-step workflow that prepares context, designs ablations, reviews feasibility, and implements runs. This approach ensures claims are justified with systematic, paper-ready results.

Yes, you can generate an EXPERIMENT_LOG.md update for traceability. The structured ablation planning process automatically outputs documentation that logs component removals, hyperparameter variations, and run orders to maintain a clear experiment history.

No specific frameworks are required to run an ablation workflow as there are no external dependencies. You provide the research context and component lists, and the system outputs a structured plan and executable workflow for your experiments.