Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 61.9048 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 78.0000 ± 0.0000
average_execution_time: 0.9068 ± 0.3551
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 77.0000 ± 0.0000
    exact_code_matches: 35.0000 ± 0.0000
    exact_code_match_rate: 29.1667 ± 0.0000
    pv_exact_matched: 77.0000 ± 0.0000
    pv_exact_mismatched: 43.0000 ± 0.0000
    pv_exact_match_rate: 64.1667 ± 0.0000
    average_pv_match_rate: 70.0000 ± 0.0000
    average_pv_mismatch_rate: 30.0000 ± 0.0000
    timing_matched: 77.0000 ± 0.0000
    timing_mismatched: 43.0000 ± 0.0000
    timing_match_rate: 64.1667 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 77.0000 ± 0.0000
    full_mismatched: 43.0000 ± 0.0000
    full_match_rate: 64.1667 ± 0.0000
    accuracy: 64.1667 ± 0.0000
    average_timing_score: 0.6391 ± 0.0000
    average_temp_score: 0.9083 ± 0.0000
    average_full_score: 0.6878 ± 0.0000
    average_best_codebleu:
      codebleu: 0.4606 ± 0.0000
      ngram_match_score: 0.0999 ± 0.0000
      weighted_ngram_match_score: 0.0615 ± 0.0000
      syntax_match_score: 0.6809 ± 0.0000
      dataflow_match_score: 0.1000 ± 0.0000
    average_best_levenshtein: 10.4333 ± 0.0000
    average_best_normalized_levenshtein: 0.3107 ± 0.0000
    average_inference_time: 0.8926 ± 0.3572
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 1.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 1.0000 ± 0.0000
    pv_exact_mismatched: 5.0000 ± 0.0000
    pv_exact_match_rate: 16.6667 ± 0.0000
    average_pv_match_rate: 54.4444 ± 0.0000
    average_pv_mismatch_rate: 45.5556 ± 0.0000
    timing_matched: 1.0000 ± 0.0000
    timing_mismatched: 5.0000 ± 0.0000
    timing_match_rate: 16.6667 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 1.0000 ± 0.0000
    full_mismatched: 5.0000 ± 0.0000
    full_match_rate: 16.6667 ± 0.0000
    accuracy: 16.6667 ± 0.0000
    average_timing_score: 0.1651 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.4686 ± 0.0000
    average_best_codebleu:
      codebleu: 0.6493 ± 0.0000
      ngram_match_score: 0.5524 ± 0.0000
      weighted_ngram_match_score: 0.5739 ± 0.0000
      syntax_match_score: 0.6958 ± 0.0000
      dataflow_match_score: 0.7749 ± 0.0000
    average_best_levenshtein: 31.8333 ± 0.0000
    average_best_normalized_levenshtein: 0.2245 ± 0.0000
    average_inference_time: 1.1899 ± 0.3126
==================================================
