Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 84.1270 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 106.0000 ± 0.0000
average_execution_time: 0.8507 ± 0.1046
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 102.0000 ± 0.0000
    exact_code_matches: 93.0000 ± 0.0000
    exact_code_match_rate: 77.5000 ± 0.0000
    pv_exact_matched: 103.0000 ± 0.0000
    pv_exact_mismatched: 17.0000 ± 0.0000
    pv_exact_match_rate: 85.8333 ± 0.0000
    average_pv_match_rate: 85.8333 ± 0.0000
    average_pv_mismatch_rate: 14.1667 ± 0.0000
    timing_matched: 102.0000 ± 0.0000
    timing_mismatched: 18.0000 ± 0.0000
    timing_match_rate: 85.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 102.0000 ± 0.0000
    full_mismatched: 18.0000 ± 0.0000
    full_match_rate: 85.0000 ± 0.0000
    accuracy: 85.0000 ± 0.0000
    average_timing_score: 0.8542 ± 0.0000
    average_temp_score: 0.9500 ± 0.0000
    average_full_score: 0.8575 ± 0.0000
    average_best_codebleu:
      codebleu: 0.5301 ± 0.0000
      ngram_match_score: 0.1698 ± 0.0000
      weighted_ngram_match_score: 0.0740 ± 0.0000
      syntax_match_score: 0.8767 ± 0.0000
      dataflow_match_score: 0.0917 ± 0.0000
    average_best_levenshtein: 3.5000 ± 0.0000
    average_best_normalized_levenshtein: 0.1012 ± 0.0000
    average_inference_time: 0.7696 ± 0.0676
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 4.0000 ± 0.0000
    exact_code_matches: 2.0000 ± 0.0000
    exact_code_match_rate: 33.3333 ± 0.0000
    pv_exact_matched: 4.0000 ± 0.0000
    pv_exact_mismatched: 2.0000 ± 0.0000
    pv_exact_match_rate: 66.6667 ± 0.0000
    average_pv_match_rate: 75.1244 ± 0.0000
    average_pv_mismatch_rate: 24.8756 ± 0.0000
    timing_matched: 4.0000 ± 0.0000
    timing_mismatched: 2.0000 ± 0.0000
    timing_match_rate: 66.6667 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 4.0000 ± 0.0000
    full_mismatched: 2.0000 ± 0.0000
    full_match_rate: 66.6667 ± 0.0000
    accuracy: 66.6667 ± 0.0000
    average_timing_score: 0.6653 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.7341 ± 0.0000
    average_best_codebleu:
      codebleu: 0.7944 ± 0.0000
      ngram_match_score: 0.6291 ± 0.0000
      weighted_ngram_match_score: 0.6584 ± 0.0000
      syntax_match_score: 0.9540 ± 0.0000
      dataflow_match_score: 0.9359 ± 0.0000
    average_best_levenshtein: 14.0000 ± 0.0000
    average_best_normalized_levenshtein: 0.0713 ± 0.0000
    average_inference_time: 2.4715 ± 0.8467
==================================================
