Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 81.6176 ± 0.0000
total_entries: 136.0000 ± 0.0000
correct_matches: 111.0000 ± 0.0000
average_execution_time: 1.0859 ± 0.0787
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 106.0000 ± 0.0000
    exact_code_matches: 94.0000 ± 0.0000
    exact_code_match_rate: 81.0345 ± 0.0000
    pv_exact_matched: 106.0000 ± 0.0000
    pv_exact_mismatched: 10.0000 ± 0.0000
    pv_exact_match_rate: 91.3793 ± 0.0000
    average_pv_match_rate: 94.5115 ± 0.0000
    average_pv_mismatch_rate: 5.4885 ± 0.0000
    timing_matched: 113.0000 ± 0.0000
    timing_mismatched: 3.0000 ± 0.0000
    timing_match_rate: 97.4138 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 106.0000 ± 0.0000
    full_mismatched: 10.0000 ± 0.0000
    full_match_rate: 91.3793 ± 0.0000
    accuracy: 91.3793 ± 0.0000
    average_timing_score: 0.9797 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9520 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5897 ± 0.0000
      comb_7_CodeBLEU: 0.8322 ± 0.0000
      dataflow_match_score: 0.2328 ± 0.0000
      ngram_match_score: 0.2289 ± 0.0000
      syntax_match_score: 0.9878 ± 0.0000
      weighted_ngram_match_score: 0.1420 ± 0.0000
    average_best_levenshtein: 1.8448 ± 0.0000
    average_best_normalized_levenshtein: 0.0621 ± 0.0000
    average_inference_time: 1.0622 ± 0.0412
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 5.0000 ± 0.0000
    exact_code_matches: 0.6667 ± 0.5774
    exact_code_match_rate: 3.3333 ± 2.8868
    pv_exact_matched: 7.0000 ± 0.0000
    pv_exact_mismatched: 13.0000 ± 0.0000
    pv_exact_match_rate: 35.0000 ± 0.0000
    average_pv_match_rate: 61.4668 ± 1.2810
    average_pv_mismatch_rate: 38.5332 ± 1.2810
    timing_matched: 12.0000 ± 0.0000
    timing_mismatched: 8.0000 ± 0.0000
    timing_match_rate: 60.0000 ± 0.0000
    temp_matched: 2.0000 ± 0.0000
    temp_mismatched: 7.0000 ± 0.0000
    temp_match_rate: 22.2222 ± 0.0000
    full_matched: 5.0000 ± 0.0000
    full_mismatched: 15.0000 ± 0.0000
    full_match_rate: 25.0000 ± 0.0000
    accuracy: 25.0000 ± 0.0000
    average_timing_score: 0.7864 ± 0.0204
    average_temp_score: 0.7080 ± 0.0088
    average_full_score: 0.6238 ± 0.0089
    average_best_codebleu:
      average_CodeBLEU: 0.3382 ± 0.0074
      comb_7_CodeBLEU: 0.4714 ± 0.0107
      dataflow_match_score: 0.4240 ± 0.0116
      ngram_match_score: 0.1086 ± 0.0031
      syntax_match_score: 0.5632 ± 0.0075
      weighted_ngram_match_score: 0.1235 ± 0.0028
    average_best_levenshtein: 154.8167 ± 2.3240
    average_best_normalized_levenshtein: 0.5069 ± 0.0054
    average_inference_time: 1.2231 ± 0.3067
==================================================
