The benchmark HTML output previously displayed abstract "score" values and scaled bars by the highest score. This change replaces scores with the actual minimum execution time in seconds (e.g., "0.15s") and scales bar widths proportionally to the highest time, making the charts directly interpretable as performance comparisons. The `run_bench` script now computes `time` from `min(result["times"])` instead of `get_score()`, and the static performance documentation is updated to reflect the new time-based values.