Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 84.9206 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 107.0000 ± 0.0000
average_execution_time: 0.8937 ± 0.1976
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 104.0000 ± 0.0000
    exact_code_matches: 96.0000 ± 0.0000
    exact_code_match_rate: 80.0000 ± 0.0000
    pv_exact_matched: 104.0000 ± 0.0000
    pv_exact_mismatched: 16.0000 ± 0.0000
    pv_exact_match_rate: 86.6667 ± 0.0000
    average_pv_match_rate: 93.3333 ± 0.0000
    average_pv_mismatch_rate: 6.6667 ± 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: 104.0000 ± 0.0000
    full_mismatched: 16.0000 ± 0.0000
    full_match_rate: 86.6667 ± 0.0000
    accuracy: 86.6667 ± 0.0000
    average_timing_score: 0.9996 ± 0.0004
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9466 ± 0.0001
    average_best_codebleu:
      average_CodeBLEU: 0.5896 ± 0.0000
      comb_7_CodeBLEU: 0.8276 ± 0.0000
      dataflow_match_score: 0.1667 ± 0.0000
      ngram_match_score: 0.2352 ± 0.0000
      syntax_match_score: 0.9725 ± 0.0000
      weighted_ngram_match_score: 0.1509 ± 0.0000
    average_best_levenshtein: 4.7250 ± 0.0000
    average_best_normalized_levenshtein: 0.0698 ± 0.0000
    average_inference_time: 0.7724 ± 0.1129
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 3.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 3.0000 ± 0.0000
    pv_exact_mismatched: 3.0000 ± 0.0000
    pv_exact_match_rate: 50.0000 ± 0.0000
    average_pv_match_rate: 70.6782 ± 0.0000
    average_pv_mismatch_rate: 29.3218 ± 0.0000
    timing_matched: 5.0000 ± 0.0000
    timing_mismatched: 1.0000 ± 0.0000
    timing_match_rate: 83.3333 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 3.0000 ± 0.0000
    full_mismatched: 3.0000 ± 0.0000
    full_match_rate: 50.0000 ± 0.0000
    accuracy: 50.0000 ± 0.0000
    average_timing_score: 0.8457 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.7346 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5849 ± 0.0000
      comb_7_CodeBLEU: 0.6973 ± 0.0000
      dataflow_match_score: 0.7655 ± 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: 3.3205 ± 1.9168
==================================================
