experiment-narrative-analysis

To write a publication-ready narrative from ML experiment results, provide the experiment metadata including seed, dataset, metrics, and baselines. The skill converts this data into a cautious, evidence-aligned paragraph suitable for research papers.

Evidence alignment in experiment narrative analysis ensures generated results narratives faithfully reflect actual metrics, plots, and ablations. It grounds written text in observed data rather than overstating claims, maintaining cautious reporting.

Yes, you can generate ablation discussions and failure analyses from dry runs. The skill supports ML research workflows across dry runs, unit tests, ablations, and full experiments to produce paper-ready discussions and caveats.

To get figure recommendations for your research paper, input your experimental results and notes. The skill analyzes ML experiment metadata and suggests appropriate plots and figures to integrate into your paper-writing workflow.

Yes, you need experiment baselines along with seed, dataset, and metrics metadata. These inputs are required to accurately align the generated narrative with experimental evidence and produce suggested follow-up analyses.

Automated narrative analysis for ML experiments relies strictly on provided metadata; it cannot infer missing metrics or baselines. You must supply comprehensive experiment details to receive accurate caveats and limitations.