Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 24.2647 ± 0.0000
total_entries: 136.0000 ± 0.0000
correct_matches: 33.0000 ± 0.0000
average_execution_time: 236.6805 ± 0.3500
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 32.0000 ± 0.0000
    exact_code_matches: 30.0000 ± 0.0000
    exact_code_match_rate: 25.8621 ± 0.0000
    pv_exact_matched: 35.0000 ± 0.0000
    pv_exact_mismatched: 81.0000 ± 0.0000
    pv_exact_match_rate: 30.1724 ± 0.0000
    average_pv_match_rate: 33.1466 ± 0.0000
    average_pv_mismatch_rate: 66.8534 ± 0.0000
    timing_matched: 35.0000 ± 0.0000
    timing_mismatched: 81.0000 ± 0.0000
    timing_match_rate: 30.1724 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 32.0000 ± 0.0000
    full_mismatched: 84.0000 ± 0.0000
    full_match_rate: 27.5862 ± 0.0000
    accuracy: 27.5862 ± 0.0000
    average_timing_score: 0.3326 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.3317 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.4030 ± 0.0000
      comb_7_CodeBLEU: 0.5881 ± 0.0000
      dataflow_match_score: 0.1638 ± 0.0000
      ngram_match_score: 0.1048 ± 0.0000
      syntax_match_score: 0.4231 ± 0.0000
      weighted_ngram_match_score: 0.0841 ± 0.0000
    average_best_levenshtein: 81.2759 ± 0.0000
    average_best_normalized_levenshtein: 0.5981 ± 0.0000
    average_inference_time: 240.7601 ± 0.3138
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 1.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 2.0000 ± 0.0000
    pv_exact_mismatched: 18.0000 ± 0.0000
    pv_exact_match_rate: 10.0000 ± 0.0000
    average_pv_match_rate: 30.9853 ± 0.0000
    average_pv_mismatch_rate: 69.0147 ± 0.0000
    timing_matched: 4.0000 ± 0.0000
    timing_mismatched: 16.0000 ± 0.0000
    timing_match_rate: 20.0000 ± 0.0000
    temp_matched: 5.0000 ± 0.0000
    temp_mismatched: 4.0000 ± 0.0000
    temp_match_rate: 55.5556 ± 0.0000
    full_matched: 1.0000 ± 0.0000
    full_mismatched: 19.0000 ± 0.0000
    full_match_rate: 5.0000 ± 0.0000
    accuracy: 5.0000 ± 0.0000
    average_timing_score: 0.3082 ± 0.0000
    average_temp_score: 0.8320 ± 0.0000
    average_full_score: 0.3239 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.3466 ± 0.0000
      comb_7_CodeBLEU: 0.4848 ± 0.0000
      dataflow_match_score: 0.2506 ± 0.0000
      ngram_match_score: 0.1151 ± 0.0000
      syntax_match_score: 0.3033 ± 0.0000
      weighted_ngram_match_score: 0.1174 ± 0.0000
    average_best_levenshtein: 191.5500 ± 0.0000
    average_best_normalized_levenshtein: 0.7216 ± 0.0000
    average_inference_time: 213.0194 ± 0.5734
==================================================
