analyze-results

To analyze ML experiment results across multiple runs, point the Skill to your results directory. It parses output files to compute mean/std statistics, delta vs baseline, and reproducibility checks across models and seeds.

The best way to compare ML model performance is by building a structured comparison table. This Skill parses result files to organize performance metrics by model, hyperparameters, and data config for direct comparison.

Yes, you can check reproducibility and flag outliers in your ML experiments. The Skill computes statistics across multiple seeds, applies reproducibility assessments, and automatically identifies performance trends or outlier runs.

To generate a comparison report from your figures and results directories, simply provide the directory path. The Skill locates and parses result files to produce structured statistics, concise findings, and recommended next experiments.

Yes, this ML results analysis tool recommends what experiments to run next. After computing statistics and identifying trends across your runs, it generates concise findings and proposes actionable next experiments.