Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 58.2011 ± 0.4582
total_entries: 126.0000 ± 0.0000
correct_matches: 73.3333 ± 0.5774
average_execution_time: 0.7629 ± 0.0988
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 71.3333 ± 0.5774
    exact_code_matches: 35.0000 ± 0.0000
    exact_code_match_rate: 29.1667 ± 0.0000
    pv_exact_matched: 71.3333 ± 0.5774
    pv_exact_mismatched: 48.6667 ± 0.5774
    pv_exact_match_rate: 59.4444 ± 0.4811
    average_pv_match_rate: 68.4722 ± 0.2406
    average_pv_mismatch_rate: 31.5278 ± 0.2406
    timing_matched: 93.0000 ± 0.0000
    timing_mismatched: 27.0000 ± 0.0000
    timing_match_rate: 77.5000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 71.3333 ± 0.5774
    full_mismatched: 48.6667 ± 0.5774
    full_match_rate: 59.4444 ± 0.4811
    accuracy: 59.4444 ± 0.4811
    average_timing_score: 0.7733 ± 0.0000
    average_temp_score: 0.9083 ± 0.0000
    average_full_score: 0.7024 ± 0.0019
    average_best_codebleu:
      average_CodeBLEU: 0.4606 ± 0.0000
      comb_7_CodeBLEU: 0.6885 ± 0.0000
      dataflow_match_score: 0.1000 ± 0.0000
      ngram_match_score: 0.0999 ± 0.0000
      syntax_match_score: 0.6809 ± 0.0000
      weighted_ngram_match_score: 0.0615 ± 0.0000
    average_best_levenshtein: 10.4333 ± 0.0000
    average_best_normalized_levenshtein: 0.3107 ± 0.0000
    average_inference_time: 0.7174 ± 0.0447
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 2.0000 ± 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: 4.0000 ± 0.0000
    pv_exact_match_rate: 33.3333 ± 0.0000
    average_pv_match_rate: 71.1111 ± 0.0000
    average_pv_mismatch_rate: 28.8889 ± 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: 2.0000 ± 0.0000
    full_mismatched: 4.0000 ± 0.0000
    full_match_rate: 33.3333 ± 0.0000
    accuracy: 33.3333 ± 0.0000
    average_timing_score: 0.8070 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.7303 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5996 ± 0.0000
      comb_7_CodeBLEU: 0.7110 ± 0.0000
      dataflow_match_score: 0.7714 ± 0.0000
      ngram_match_score: 0.4049 ± 0.0000
      syntax_match_score: 0.7990 ± 0.0000
      weighted_ngram_match_score: 0.4231 ± 0.0000
    average_best_levenshtein: 32.1667 ± 0.5774
    average_best_normalized_levenshtein: 0.2261 ± 0.0027
    average_inference_time: 1.6727 ± 1.1885
==================================================
