Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 48.9418 ± 0.9164
total_entries: 126.0000 ± 0.0000
correct_matches: 61.6667 ± 1.1547
average_execution_time: 0.9456 ± 0.0286
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 61.6667 ± 1.1547
    exact_code_matches: 30.0000 ± 0.0000
    exact_code_match_rate: 25.0000 ± 0.0000
    pv_exact_matched: 61.6667 ± 1.1547
    pv_exact_mismatched: 58.3333 ± 1.1547
    pv_exact_match_rate: 51.3889 ± 0.9623
    average_pv_match_rate: 61.1806 ± 0.4811
    average_pv_mismatch_rate: 38.8194 ± 0.4811
    timing_matched: 87.0000 ± 0.0000
    timing_mismatched: 33.0000 ± 0.0000
    timing_match_rate: 72.5000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 61.6667 ± 1.1547
    full_mismatched: 58.3333 ± 1.1547
    full_match_rate: 51.3889 ± 0.9623
    accuracy: 51.3889 ± 0.9623
    average_timing_score: 0.7235 ± 0.0000
    average_temp_score: 0.9083 ± 0.0000
    average_full_score: 0.6341 ± 0.0038
    average_best_codebleu:
      average_CodeBLEU: 0.4870 ± 0.0000
      comb_7_CodeBLEU: 0.7328 ± 0.0000
      dataflow_match_score: 0.1083 ± 0.0000
      ngram_match_score: 0.0823 ± 0.0000
      syntax_match_score: 0.7934 ± 0.0000
      weighted_ngram_match_score: 0.0724 ± 0.0000
    average_best_levenshtein: 68.9528 ± 0.0385
    average_best_normalized_levenshtein: 0.3873 ± 0.0001
    average_inference_time: 0.9136 ± 0.0230
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 0.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 0.0000 ± 0.0000
    pv_exact_mismatched: 6.0000 ± 0.0000
    pv_exact_match_rate: 0.0000 ± 0.0000
    average_pv_match_rate: 7.6599 ± 0.0000
    average_pv_mismatch_rate: 92.3401 ± 0.0000
    timing_matched: 0.0000 ± 0.0000
    timing_mismatched: 6.0000 ± 0.0000
    timing_match_rate: 0.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 0.0000 ± 0.0000
    full_mismatched: 6.0000 ± 0.0000
    full_match_rate: 0.0000 ± 0.0000
    accuracy: 0.0000 ± 0.0000
    average_timing_score: 0.1422 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.0897 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.3318 ± 0.0041
      comb_7_CodeBLEU: 0.4765 ± 0.0050
      dataflow_match_score: 0.5170 ± 0.0004
      ngram_match_score: 0.0810 ± 0.0021
      syntax_match_score: 0.4621 ± 0.0107
      weighted_ngram_match_score: 0.1003 ± 0.0031
    average_best_levenshtein: 125.0556 ± 5.2924
    average_best_normalized_levenshtein: 0.6450 ± 0.0120
    average_inference_time: 1.5858 ± 0.4028
==================================================
