evaluation · metrics

Live view of eval/run_eval.py output + current corpus state. All numbers are real — re-run the harness and refresh to see new values.