| |
| from __future__ import annotations |
| import argparse, csv |
| from pathlib import Path |
|
|
| import sys |
| ROOT = Path(__file__).resolve().parents[1] |
| if str(ROOT) not in sys.path: |
| sys.path.insert(0, str(ROOT)) |
| from statistics import mean |
| from tbg_cot_bench.core import load_scenarios |
| from tbg_cot_bench.converter_baseline import EvidenceConverter |
|
|
|
|
| def main() -> None: |
| parser = argparse.ArgumentParser(description="Evaluate baseline converter against gold scenario labels.") |
| parser.add_argument("--scenarios", default="scenarios") |
| parser.add_argument("--out", default="results") |
| args = parser.parse_args() |
| out = Path(args.out); out.mkdir(parents=True, exist_ok=True) |
|
|
| converter = EvidenceConverter() |
| rows = [] |
| for scenario in load_scenarios(args.scenarios): |
| event_a = scenario["events"]["event_a"]["label"] |
| event_b = scenario["events"]["event_b"]["label"] |
| parsed = converter.convert_steps([s["text"] for s in scenario["steps"]], scenario_id=scenario["id"], event_a_label=event_a, event_b_label=event_b) |
| for idx, (gold, pred) in enumerate(zip(scenario["steps"], parsed), start=1): |
| pred_dict = pred.to_dict() |
| rows.append({ |
| "scenario_id": scenario["id"], |
| "step": idx, |
| "gold_supports_forward": gold["supports_forward"], |
| "pred_supports_forward": pred.supports_forward, |
| "direction_correct": pred.supports_forward == gold["supports_forward"], |
| "gold_strength": gold["strength"], |
| "pred_strength": pred.strength, |
| "strength_abs_error": round(abs(float(gold["strength"]) - pred.strength), 4), |
| "gold_source": gold["source"], |
| "pred_source": pred.source_label, |
| "gold_source_weight": gold["source_weight"], |
| "pred_source_weight": pred.source_weight, |
| "source_weight_abs_error": round(abs(float(gold["source_weight"]) - pred.source_weight), 4), |
| "matched_rule": pred_dict["meta"]["matched_rule"], |
| "notes": "|".join(pred_dict["meta"]["notes"]), |
| }) |
|
|
| fieldnames = list(rows[0].keys()) |
| with (out / "converter_eval.csv").open("w", newline="", encoding="utf-8") as f: |
| writer = csv.DictWriter(f, fieldnames=fieldnames) |
| writer.writeheader(); writer.writerows(rows) |
|
|
| acc = mean(1.0 if r["direction_correct"] else 0.0 for r in rows) |
| strength_mae = mean(float(r["strength_abs_error"]) for r in rows) |
| weight_mae = mean(float(r["source_weight_abs_error"]) for r in rows) |
| summary = [ |
| {"metric": "direction_accuracy", "value": round(acc, 4)}, |
| {"metric": "strength_mae", "value": round(strength_mae, 4)}, |
| {"metric": "source_weight_mae", "value": round(weight_mae, 4)}, |
| {"metric": "num_steps", "value": len(rows)}, |
| ] |
| with (out / "converter_eval_summary.csv").open("w", newline="", encoding="utf-8") as f: |
| writer = csv.DictWriter(f, fieldnames=["metric", "value"]) |
| writer.writeheader(); writer.writerows(summary) |
| print(f"Direction accuracy: {acc:.3f}; wrote {out / 'converter_eval.csv'}") |
|
|
| if __name__ == "__main__": |
| main() |
|
|
