Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 90.6863 ± 0.4245
total_entries: 136.0000 ± 0.0000
correct_matches: 123.3333 ± 0.5774
average_execution_time: 1.7377 ± 0.1315
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 115.0000 ± 0.0000
    exact_code_matches: 110.0000 ± 0.0000
    exact_code_match_rate: 94.8276 ± 0.0000
    pv_exact_matched: 115.0000 ± 0.0000
    pv_exact_mismatched: 1.0000 ± 0.0000
    pv_exact_match_rate: 99.1379 ± 0.0000
    average_pv_match_rate: 99.7126 ± 0.0000
    average_pv_mismatch_rate: 0.2874 ± 0.0000
    timing_matched: 115.0000 ± 0.0000
    timing_mismatched: 1.0000 ± 0.0000
    timing_match_rate: 99.1379 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 115.0000 ± 0.0000
    full_mismatched: 1.0000 ± 0.0000
    full_match_rate: 99.1379 ± 0.0000
    accuracy: 99.1379 ± 0.0000
    average_timing_score: 0.9953 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9968 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.6059 ± 0.0000
      comb_7_CodeBLEU: 0.8421 ± 0.0000
      dataflow_match_score: 0.2328 ± 0.0000
      ngram_match_score: 0.2626 ± 0.0000
      syntax_match_score: 0.9990 ± 0.0000
      weighted_ngram_match_score: 0.1620 ± 0.0000
    average_best_levenshtein: 0.2845 ± 0.0000
    average_best_normalized_levenshtein: 0.0101 ± 0.0000
    average_inference_time: 0.7948 ± 0.1440
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 8.3333 ± 0.5774
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 10.0000 ± 0.0000
    pv_exact_mismatched: 10.0000 ± 0.0000
    pv_exact_match_rate: 50.0000 ± 0.0000
    average_pv_match_rate: 73.2463 ± 0.0000
    average_pv_mismatch_rate: 26.7537 ± 0.0000
    timing_matched: 10.3333 ± 0.5774
    timing_mismatched: 9.6667 ± 0.5774
    timing_match_rate: 51.6667 ± 2.8868
    temp_matched: 6.0000 ± 0.0000
    temp_mismatched: 3.0000 ± 0.0000
    temp_match_rate: 66.6667 ± 0.0000
    full_matched: 8.3333 ± 0.5774
    full_mismatched: 11.6667 ± 0.5774
    full_match_rate: 41.6667 ± 2.8868
    accuracy: 41.6667 ± 2.8868
    average_timing_score: 0.7014 ± 0.0110
    average_temp_score: 0.9065 ± 0.0000
    average_full_score: 0.7217 ± 0.0022
    average_best_codebleu:
      average_CodeBLEU: 0.4108 ± 0.0010
      comb_7_CodeBLEU: 0.5303 ± 0.0004
      dataflow_match_score: 0.5105 ± 0.0000
      ngram_match_score: 0.1934 ± 0.0020
      syntax_match_score: 0.6093 ± 0.0000
      weighted_ngram_match_score: 0.2300 ± 0.0018
    average_best_levenshtein: 151.2500 ± 0.1732
    average_best_normalized_levenshtein: 0.4534 ± 0.0023
    average_inference_time: 7.2068 ± 0.1441
==================================================
