| name | experiment-audit |
| description | Audit experiments for top ML, CV, systems, and architecture submissions, checking baselines, ablations, metrics, statistical strength, datasets, stress tests, failure cases, and claim coverage. |
| version | 0.1.0 |
Experiment Audit
Use this skill before submission when the evidence might be thin.
Workflow
- Extract each major paper claim.
- Map claims to experiments and identify missing support.
- Check baselines, ablations, metrics, seeds, splits, compute, runtime,
qualitative examples, and failure cases.
- Prioritize new experiments by reviewer impact and time cost.
Output
Return Claim-evidence table, Missing baselines, Ablation gaps,
High-impact experiments, and Reviewer risk.