Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 74.6032 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 94.0000 ± 0.0000
average_execution_time: 1.5609 ± 0.0575
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 92.0000 ± 0.0000
    exact_code_matches: 80.0000 ± 0.0000
    exact_code_match_rate: 66.6667 ± 0.0000
    pv_exact_matched: 92.0000 ± 0.0000
    pv_exact_mismatched: 28.0000 ± 0.0000
    pv_exact_match_rate: 76.6667 ± 0.0000
    average_pv_match_rate: 76.6667 ± 0.0000
    average_pv_mismatch_rate: 23.3333 ± 0.0000
    timing_matched: 92.0000 ± 0.0000
    timing_mismatched: 28.0000 ± 0.0000
    timing_match_rate: 76.6667 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 92.0000 ± 0.0000
    full_mismatched: 28.0000 ± 0.0000
    full_match_rate: 76.6667 ± 0.0000
    accuracy: 76.6667 ± 0.0000
    average_timing_score: 0.7667 ± 0.0000
    average_temp_score: 0.9083 ± 0.0000
    average_full_score: 0.7667 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5364 ± 0.0000
      comb_7_CodeBLEU: 0.7906 ± 0.0000
      dataflow_match_score: 0.1000 ± 0.0000
      ngram_match_score: 0.1529 ± 0.0000
      syntax_match_score: 0.9200 ± 0.0000
      weighted_ngram_match_score: 0.0726 ± 0.0000
    average_best_levenshtein: 3.4917 ± 0.0000
    average_best_normalized_levenshtein: 0.1208 ± 0.0000
    average_inference_time: 1.4186 ± 0.0308
  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: 33.3333 ± 0.0000
    average_pv_mismatch_rate: 66.6667 ± 0.0000
    timing_matched: 2.0000 ± 0.0000
    timing_mismatched: 4.0000 ± 0.0000
    timing_match_rate: 33.3333 ± 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: 0.3321 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.3331 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.6915 ± 0.0000
      comb_7_CodeBLEU: 0.8033 ± 0.0000
      dataflow_match_score: 0.8894 ± 0.0000
      ngram_match_score: 0.4617 ± 0.0000
      syntax_match_score: 0.8663 ± 0.0000
      weighted_ngram_match_score: 0.5484 ± 0.0000
    average_best_levenshtein: 31.6667 ± 0.0000
    average_best_normalized_levenshtein: 0.1664 ± 0.0000
    average_inference_time: 4.4080 ± 0.5987
==================================================
