analyze-results

To analyze ML experiment results, the Skill parses JSON and CSV files from results or figures directories, aggregates metrics by model and configuration, and computes statistics like means and standard deviations to summarize performance.

Yes, you can compare ML models across multiple seeds. The Skill aggregates metrics by model and configuration, computes delta versus baseline, assesses reproducibility, and generates structured comparison tables highlighting trends and outliers.

The best way to generate insights from hyperparameter sweeps is to parse result files, aggregate metrics by configuration, and produce structured narrative findings with observations, interpretations, and implications to guide next steps.

Yes, this ML results analysis tool works without external dependencies. It independently parses result files in JSON and CSV formats across your directories to compute statistics and surface actionable insights.

To assess reproducibility and outliers in machine learning experiments, the Skill computes statistics across multiple seeds, evaluates delta versus baseline, and builds comparison tables to highlight trends and anomalies.

You can compute statistics such as means, standard deviations, and p-values from your machine learning experiment results. The Skill summarizes performance across models and seeds to support decision-making.