Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 80.9524 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 102.0000 ± 0.0000
average_execution_time: 1.0526 ± 0.1249
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 101.0000 ± 0.0000
    exact_code_matches: 92.0000 ± 0.0000
    exact_code_match_rate: 76.6667 ± 0.0000
    pv_exact_matched: 101.0000 ± 0.0000
    pv_exact_mismatched: 19.0000 ± 0.0000
    pv_exact_match_rate: 84.1667 ± 0.0000
    average_pv_match_rate: 91.3889 ± 0.0000
    average_pv_mismatch_rate: 8.6111 ± 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: 101.0000 ± 0.0000
    full_mismatched: 19.0000 ± 0.0000
    full_match_rate: 84.1667 ± 0.0000
    accuracy: 84.1667 ± 0.0000
    average_timing_score: 1.0000 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9311 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5707 ± 0.0000
      comb_7_CodeBLEU: 0.8214 ± 0.0000
      dataflow_match_score: 0.1000 ± 0.0000
      ngram_match_score: 0.1984 ± 0.0000
      syntax_match_score: 0.9769 ± 0.0000
      weighted_ngram_match_score: 0.1077 ± 0.0000
    average_best_levenshtein: 3.5167 ± 0.0000
    average_best_normalized_levenshtein: 0.0832 ± 0.0000
    average_inference_time: 1.0357 ± 0.0973
  complex:
    count: 6.0000 ± 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: 1.0000 ± 0.0000
    pv_exact_mismatched: 5.0000 ± 0.0000
    pv_exact_match_rate: 16.6667 ± 0.0000
    average_pv_match_rate: 72.8315 ± 0.0000
    average_pv_mismatch_rate: 27.1685 ± 0.0000
    timing_matched: 6.0000 ± 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: 1.0000 ± 0.0000
    full_mismatched: 5.0000 ± 0.0000
    full_match_rate: 16.6667 ± 0.0000
    accuracy: 16.6667 ± 0.0000
    average_timing_score: 0.8943 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.7615 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5393 ± 0.0001
      comb_7_CodeBLEU: 0.6823 ± 0.0000
      dataflow_match_score: 0.5737 ± 0.0000
      ngram_match_score: 0.2766 ± 0.0001
      syntax_match_score: 0.6482 ± 0.0000
      weighted_ngram_match_score: 0.3255 ± 0.0002
    average_best_levenshtein: 79.2222 ± 3.3679
    average_best_normalized_levenshtein: 0.4733 ± 0.0037
    average_inference_time: 1.3905 ± 0.8072
==================================================
