examples/evaluate_track_a.py

from __future__ import annotations

import argparse
import json
from pathlib import Path

from datasets import load_dataset


def main() -> None:
    parser = argparse.ArgumentParser(
        description="Score RoutingBench Track A predictions."
    )
    parser.add_argument("predictions", type=Path)
    parser.add_argument("--split", choices=("development", "validation", "test"))
    parser.add_argument("--revision", default=None)
    args = parser.parse_args()

    rows = load_dataset(
        "TrangBui/RoutingBench-V2",
        "decision",
        split=args.split,
        revision=args.revision,
    )
    gold = {row["decision_id"]: set(row["admissible_actions"]) for row in rows}
    predictions = {}
    with args.predictions.open("r", encoding="utf-8") as handle:
        for line in handle:
            item = json.loads(line)
            predictions[str(item["decision_id"])] = str(item["selected_action"])

    missing = sorted(set(gold) - set(predictions))
    unknown = sorted(set(predictions) - set(gold))
    if missing or unknown:
        raise SystemExit(
            f"Prediction IDs mismatch: missing={len(missing)}, unknown={len(unknown)}"
        )
    correct = sum(predictions[key] in actions for key, actions in gold.items())
    print(
        json.dumps(
            {
                "split": args.split,
                "count": len(gold),
                "admissible_next_action_accuracy": correct / len(gold),
            },
            indent=2,
        )
    )


if __name__ == "__main__":
    main()