Function write_evaluation

experiments/baselines/nl_profile/runner.py:583–617

(output_dir: Path, rows: List[Dict[str, Any]], episode_rows: List[Dict[str, Any]])

def write_evaluation(output_dir: Path, rows: List[Dict[str, Any]], episode_rows: List[Dict[str, Any]]) -> Dict[str, Any]:
    grouped = group_by_episode(rows)
    episode_metrics = {
        episode_id: eval_metrics.evaluate_episode(episode_group, [5, 10, 20])
        for episode_id, episode_group in grouped.items()
        if episode_id
    }
    summary = eval_metrics.aggregate_metrics(episode_metrics, [5, 10, 20])
    dataset_summary = eval_metrics.build_dataset_summary(rows, episode_rows)
    result = {
        "summary": summary,
        "dataset_summary": dataset_summary,
        "episodes": episode_metrics,
    }
    (output_dir / "evaluation_metrics.json").write_text(
        json.dumps(result, ensure_ascii=False, indent=2),
        encoding="utf-8",
    )
    (output_dir / "dataset_summary.json").write_text(
        json.dumps(dataset_summary, ensure_ascii=False, indent=2),
        encoding="utf-8",
    )
    (output_dir / "dataset_summary.md").write_text(
        eval_metrics.build_dataset_summary_markdown(dataset_summary),
        encoding="utf-8",
    )
    (output_dir / "main_experiment_table_top20.md").write_text(
        eval_metrics.build_main_experiment_table(summary, METHOD_NAME),
        encoding="utf-8",
    )
    (output_dir / "case_metrics_table_top20.md").write_text(
        eval_metrics.build_case_metrics_table(summary, METHOD_NAME),
        encoding="utf-8",
    )
    return result


def run_baseline(