Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 87.5000 ± 0.0000
total_entries: 136.0000 ± 0.0000
correct_matches: 119.0000 ± 0.0000
average_execution_time: 1.3517 ± 0.1634
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 112.0000 ± 0.0000
    exact_code_matches: 105.0000 ± 0.0000
    exact_code_match_rate: 90.5172 ± 0.0000
    pv_exact_matched: 114.0000 ± 0.0000
    pv_exact_mismatched: 2.0000 ± 0.0000
    pv_exact_match_rate: 98.2759 ± 0.0000
    average_pv_match_rate: 98.2759 ± 0.0000
    average_pv_mismatch_rate: 1.7241 ± 0.0000
    timing_matched: 112.0000 ± 0.0000
    timing_mismatched: 4.0000 ± 0.0000
    timing_match_rate: 96.5517 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 112.0000 ± 0.0000
    full_mismatched: 4.0000 ± 0.0000
    full_match_rate: 96.5517 ± 0.0000
    accuracy: 96.5517 ± 0.0000
    average_timing_score: 0.9739 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9810 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5841 ± 0.0000
      comb_7_CodeBLEU: 0.8277 ± 0.0000
      dataflow_match_score: 0.2328 ± 0.0000
      ngram_match_score: 0.2250 ± 0.0000
      syntax_match_score: 0.9802 ± 0.0000
      weighted_ngram_match_score: 0.1312 ± 0.0000
    average_best_levenshtein: 1.3966 ± 0.0000
    average_best_normalized_levenshtein: 0.0307 ± 0.0000
    average_inference_time: 0.9905 ± 0.0449
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 7.0000 ± 0.0000
    exact_code_matches: 1.0000 ± 0.0000
    exact_code_match_rate: 5.0000 ± 0.0000
    pv_exact_matched: 7.3333 ± 0.5774
    pv_exact_mismatched: 12.6667 ± 0.5774
    pv_exact_match_rate: 36.6667 ± 2.8868
    average_pv_match_rate: 52.0521 ± 1.9000
    average_pv_mismatch_rate: 47.9479 ± 1.9000
    timing_matched: 14.0000 ± 0.0000
    timing_mismatched: 6.0000 ± 0.0000
    timing_match_rate: 70.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 9.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 7.0000 ± 0.0000
    full_mismatched: 13.0000 ± 0.0000
    full_match_rate: 35.0000 ± 0.0000
    accuracy: 35.0000 ± 0.0000
    average_timing_score: 0.8230 ± 0.0073
    average_temp_score: 0.5500 ± 0.0000
    average_full_score: 0.5677 ± 0.0166
    average_best_codebleu:
      average_CodeBLEU: 0.4407 ± 0.0013
      comb_7_CodeBLEU: 0.5618 ± 0.0013
      dataflow_match_score: 0.5370 ± 0.0004
      ngram_match_score: 0.2175 ± 0.0016
      syntax_match_score: 0.6482 ± 0.0029
      weighted_ngram_match_score: 0.2599 ± 0.0012
    average_best_levenshtein: 146.8667 ± 3.3198
    average_best_normalized_levenshtein: 0.4117 ± 0.0032
    average_inference_time: 3.4465 ± 1.2134
==================================================
