analyze-results

Analyzing ML experiment results involves parsing JSON and CSV files from directories like results/ or figures/ into structured data. The Skill compares metrics across models and hyperparameters, computes statistics across seeds, and generates concise summaries with recommended next steps.

The best way to compare model performance is by building structured comparison tables that highlight deltas versus a baseline. This reveals trends and outliers across hyperparameters and data configurations, pairing observations with interpretations and actionable next steps.

To check reproducibility across seeds, compute statistics like mean and standard deviation from your experiment results. The Skill parses raw output files to calculate these statistics, identify trends, detect outliers, and pair these observations with actionable interpretations.

Yes, you can automatically generate project notes by parsing raw experiment outputs into structured insights. The Skill pairs observations with interpretations and recommended next steps, preparing a concise update suitable for sharing with project stakeholders.

Yes, the tool identifies and parses JSON and CSV result files from both common output folders and project-specific directories. It loads these files into structured datasets to compare results across independent variables and compute primary metrics.