analyze-results

To analyze machine learning experiment results, parse JSON and CSV output files from your hyperparameter sweeps to compute statistics, generate comparison tables, and extract actionable insights. The tool organizes results by independent variables and highlights deltas versus a baseline.

Evaluating model performance across multiple seeds requires summarizing mean and variability statistics to surface trends. The analysis process automatically flags suspicious outliers and computes variability across multi-seed evaluations to ensure reproducibility.

Yes, baseline-versus-delta reporting works directly with JSON and CSV files. The system locates results in common project directories, parses them into structured data, and computes deltas to reveal what actually changed between your baseline and new experiments.

To structure ML experiment findings, the tool outputs numbered key findings containing observation, interpretation, implication, and next step. It also recommends documentation updates when results matter, turning raw metrics into decision-ready findings.

No, you do not need to manually format comparison tables for machine learning statistics. The tool automatically parses structured data from your results folder and generates organized tables highlighting independent variables and statistical deltas.