analyze-results

To analyze ML experiment results, this skill parses output directories like figures/ or results/, builds a comparison table by varying hyperparameters or data configurations, and computes statistics to compare model variants.

The best way to compute statistics and assess reproducibility across multiple seeds is to calculate mean and standard deviation for dependent metrics, while automatically identifying trends and outliers in the experiment outputs.

You can generate a comparison table from hyperparameter sweep outputs by locating and parsing your project-specific folders, then structuring the data by varying independent variables like model type and recording the dependent metrics.

Yes, you can extract actionable insights and recommended next steps from experiment data by performing statistical analysis on the parsed results and synthesizing the findings into report-ready summaries for your project notes.

Yes, this tool works with standard project directory structures by automatically locating and parsing results from common output directories including figures/, results/, and other project-specific folders to extract structured data.