Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 0.7353 ± 1.2736
total_entries: 136.0000 ± 0.0000
correct_matches: 1.0000 ± 1.7321
average_execution_time: 1.0608 ± 0.9540
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 1.0000 ± 1.7321
    exact_code_matches: 1.0000 ± 1.7321
    exact_code_match_rate: 0.8621 ± 1.4931
    pv_exact_matched: 1.0000 ± 1.7321
    pv_exact_mismatched: 115.0000 ± 1.7321
    pv_exact_match_rate: 0.8621 ± 1.4931
    average_pv_match_rate: 11.4224 ± 2.0904
    average_pv_mismatch_rate: 88.5776 ± 2.0904
    timing_matched: 28.0000 ± 6.9282
    timing_mismatched: 88.0000 ± 6.9282
    timing_match_rate: 24.1379 ± 5.9726
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 1.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 1.0000 ± 1.7321
    full_mismatched: 115.0000 ± 1.7321
    full_match_rate: 0.8621 ± 1.4931
    accuracy: 0.8621 ± 1.4931
    average_timing_score: 0.2414 ± 0.0597
    average_temp_score: 0.9914 ± 0.0000
    average_full_score: 0.1397 ± 0.0287
    average_best_codebleu:
      average_CodeBLEU: 0.2586 ± 0.0149
      comb_7_CodeBLEU: 0.4125 ± 0.0216
      dataflow_match_score: 0.0287 ± 0.0498
      ngram_match_score: 0.0023 ± 0.0040
      syntax_match_score: 0.0301 ± 0.0522
      weighted_ngram_match_score: 0.0020 ± 0.0035
    average_best_levenshtein: 331.2730 ± 538.0805
    average_best_normalized_levenshtein: 0.9884 ± 0.0200
    average_inference_time: 1.1558 ± 1.1186
  complex:
    count: 20.0000 ± 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: 20.0000 ± 0.0000
    pv_exact_match_rate: 0.0000 ± 0.0000
    average_pv_match_rate: 1.2500 ± 0.0000
    average_pv_mismatch_rate: 98.7500 ± 0.0000
    timing_matched: 1.0000 ± 0.0000
    timing_mismatched: 19.0000 ± 0.0000
    timing_match_rate: 5.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 9.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 0.0000 ± 0.0000
    full_mismatched: 20.0000 ± 0.0000
    full_match_rate: 0.0000 ± 0.0000
    accuracy: 0.0000 ± 0.0000
    average_timing_score: 0.0500 ± 0.0000
    average_temp_score: 0.5500 ± 0.0000
    average_full_score: 0.0200 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.2500 ± 0.0000
      comb_7_CodeBLEU: 0.4000 ± 0.0000
      dataflow_match_score: 0.0000 ± 0.0000
      ngram_match_score: 0.0000 ± 0.0000
      syntax_match_score: 0.0000 ± 0.0000
      weighted_ngram_match_score: 0.0000 ± 0.0000
    average_best_levenshtein: 256.0000 ± 0.0000
    average_best_normalized_levenshtein: 1.0000 ± 0.0000
    average_inference_time: 0.5094 ± 0.0012
==================================================
