Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 91.1765 ± 0.0000
total_entries: 136.0000 ± 0.0000
correct_matches: 124.0000 ± 0.0000
average_execution_time: 1.6395 ± 0.1850
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 116.0000 ± 0.0000
    exact_code_matches: 111.0000 ± 0.0000
    exact_code_match_rate: 95.6897 ± 0.0000
    pv_exact_matched: 116.0000 ± 0.0000
    pv_exact_mismatched: 0.0000 ± 0.0000
    pv_exact_match_rate: 100.0000 ± 0.0000
    average_pv_match_rate: 100.0000 ± 0.0000
    average_pv_mismatch_rate: 0.0000 ± 0.0000
    timing_matched: 116.0000 ± 0.0000
    timing_mismatched: 0.0000 ± 0.0000
    timing_match_rate: 100.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 116.0000 ± 0.0000
    full_mismatched: 0.0000 ± 0.0000
    full_match_rate: 100.0000 ± 0.0000
    accuracy: 100.0000 ± 0.0000
    average_timing_score: 1.0000 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 1.0000 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.6056 ± 0.0000
      comb_7_CodeBLEU: 0.8422 ± 0.0000
      dataflow_match_score: 0.2241 ± 0.0000
      ngram_match_score: 0.2620 ± 0.0000
      syntax_match_score: 1.0000 ± 0.0000
      weighted_ngram_match_score: 0.1602 ± 0.0000
    average_best_levenshtein: 0.1293 ± 0.0000
    average_best_normalized_levenshtein: 0.0079 ± 0.0000
    average_inference_time: 0.7232 ± 0.1272
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 8.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 9.0000 ± 0.0000
    pv_exact_mismatched: 11.0000 ± 0.0000
    pv_exact_match_rate: 45.0000 ± 0.0000
    average_pv_match_rate: 78.8771 ± 0.0000
    average_pv_mismatch_rate: 21.1229 ± 0.0000
    timing_matched: 11.0000 ± 0.0000
    timing_mismatched: 9.0000 ± 0.0000
    timing_match_rate: 55.0000 ± 0.0000
    temp_matched: 6.0000 ± 0.0000
    temp_mismatched: 3.0000 ± 0.0000
    temp_match_rate: 66.6667 ± 0.0000
    full_matched: 8.0000 ± 0.0000
    full_mismatched: 12.0000 ± 0.0000
    full_match_rate: 40.0000 ± 0.0000
    accuracy: 40.0000 ± 0.0000
    average_timing_score: 0.7942 ± 0.0001
    average_temp_score: 0.9517 ± 0.0000
    average_full_score: 0.7955 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.3734 ± 0.0000
      comb_7_CodeBLEU: 0.4824 ± 0.0000
      dataflow_match_score: 0.4708 ± 0.0000
      ngram_match_score: 0.1856 ± 0.0000
      syntax_match_score: 0.5892 ± 0.0000
      weighted_ngram_match_score: 0.1983 ± 0.0000
    average_best_levenshtein: 147.7000 ± 0.0000
    average_best_normalized_levenshtein: 0.4606 ± 0.0000
    average_inference_time: 6.9540 ± 0.5282
==================================================
