Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 46.0317 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 58.0000 ± 0.0000
average_execution_time: 0.9117 ± 0.0187
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 58.0000 ± 0.0000
    exact_code_matches: 30.0000 ± 0.0000
    exact_code_match_rate: 25.0000 ± 0.0000
    pv_exact_matched: 58.0000 ± 0.0000
    pv_exact_mismatched: 62.0000 ± 0.0000
    pv_exact_match_rate: 48.3333 ± 0.0000
    average_pv_match_rate: 59.5000 ± 0.0000
    average_pv_mismatch_rate: 40.5000 ± 0.0000
    timing_matched: 86.3333 ± 0.5774
    timing_mismatched: 33.6667 ± 0.5774
    timing_match_rate: 71.9444 ± 0.4811
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 58.0000 ± 0.0000
    full_mismatched: 62.0000 ± 0.0000
    full_match_rate: 48.3333 ± 0.0000
    accuracy: 48.3333 ± 0.0000
    average_timing_score: 0.7195 ± 0.0008
    average_temp_score: 0.9083 ± 0.0000
    average_full_score: 0.6199 ± 0.0002
    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: 68.9306 ± 0.0770
    average_best_normalized_levenshtein: 0.3873 ± 0.0001
    average_inference_time: 0.8678 ± 0.0282
  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: 0.1405 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.0894 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.3076 ± 0.0232
      comb_7_CodeBLEU: 0.4385 ± 0.0365
      dataflow_match_score: 0.4732 ± 0.0238
      ngram_match_score: 0.0813 ± 0.0023
      syntax_match_score: 0.4116 ± 0.0678
      weighted_ngram_match_score: 0.0979 ± 0.0038
    average_best_levenshtein: 113.3889 ± 23.3085
    average_best_normalized_levenshtein: 0.6318 ± 0.0307
    average_inference_time: 1.7909 ± 0.7151
==================================================
