Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 83.3333 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 105.0000 ± 0.0000
average_execution_time: 0.8540 ± 0.0900
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 105.0000 ± 0.0000
    exact_code_matches: 96.0000 ± 0.0000
    exact_code_match_rate: 80.0000 ± 0.0000
    pv_exact_matched: 105.0000 ± 0.0000
    pv_exact_mismatched: 15.0000 ± 0.0000
    pv_exact_match_rate: 87.5000 ± 0.0000
    average_pv_match_rate: 93.7500 ± 0.0000
    average_pv_mismatch_rate: 6.2500 ± 0.0000
    timing_matched: 120.0000 ± 0.0000
    timing_mismatched: 0.0000 ± 0.0000
    timing_match_rate: 100.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 105.0000 ± 0.0000
    full_mismatched: 15.0000 ± 0.0000
    full_match_rate: 87.5000 ± 0.0000
    accuracy: 87.5000 ± 0.0000
    average_timing_score: 0.9998 ± 0.0002
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9500 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5907 ± 0.0000
      comb_7_CodeBLEU: 0.8283 ± 0.0000
      dataflow_match_score: 0.1750 ± 0.0000
      ngram_match_score: 0.2363 ± 0.0000
      syntax_match_score: 0.9733 ± 0.0000
      weighted_ngram_match_score: 0.1533 ± 0.0000
    average_best_levenshtein: 4.7333 ± 0.0000
    average_best_normalized_levenshtein: 0.0690 ± 0.0000
    average_inference_time: 0.7838 ± 0.0940
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 0.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 0.0000 ± 0.0000
    pv_exact_mismatched: 6.0000 ± 0.0000
    pv_exact_match_rate: 0.0000 ± 0.0000
    average_pv_match_rate: 11.5385 ± 0.0000
    average_pv_mismatch_rate: 88.4615 ± 0.0000
    timing_matched: 1.0000 ± 0.0000
    timing_mismatched: 5.0000 ± 0.0000
    timing_match_rate: 16.6667 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 0.0000 ± 0.0000
    full_mismatched: 6.0000 ± 0.0000
    full_match_rate: 0.0000 ± 0.0000
    accuracy: 0.0000 ± 0.0000
    average_timing_score: 0.1367 ± 0.0000
    average_temp_score: 0.1667 ± 0.0000
    average_full_score: 0.1197 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5875 ± 0.0000
      comb_7_CodeBLEU: 0.7015 ± 0.0000
      dataflow_match_score: 0.7760 ± 0.0000
      ngram_match_score: 0.3798 ± 0.0000
      syntax_match_score: 0.7790 ± 0.0000
      weighted_ngram_match_score: 0.4154 ± 0.0000
    average_best_levenshtein: 44.1667 ± 0.0000
    average_best_normalized_levenshtein: 0.2430 ± 0.0000
    average_inference_time: 2.2576 ± 0.0103
==================================================
