Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 85.7143 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 108.0000 ± 0.0000
average_execution_time: 0.7870 ± 0.1727
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 105.0000 ± 0.0000
    exact_code_matches: 100.0000 ± 0.0000
    exact_code_match_rate: 83.3333 ± 0.0000
    pv_exact_matched: 105.0000 ± 0.0000
    pv_exact_mismatched: 15.0000 ± 0.0000
    pv_exact_match_rate: 87.5000 ± 0.0000
    average_pv_match_rate: 93.7500 ± 0.0000
    average_pv_mismatch_rate: 6.2500 ± 0.0000
    timing_matched: 120.0000 ± 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: 105.0000 ± 0.0000
    full_mismatched: 15.0000 ± 0.0000
    full_match_rate: 87.5000 ± 0.0000
    accuracy: 87.5000 ± 0.0000
    average_timing_score: 1.0000 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9500 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5753 ± 0.0000
      comb_7_CodeBLEU: 0.8228 ± 0.0000
      dataflow_match_score: 0.1000 ± 0.0000
      ngram_match_score: 0.2111 ± 0.0000
      syntax_match_score: 0.9756 ± 0.0000
      weighted_ngram_match_score: 0.1145 ± 0.0000
    average_best_levenshtein: 2.8083 ± 0.0000
    average_best_normalized_levenshtein: 0.0562 ± 0.0000
    average_inference_time: 0.6871 ± 0.1328
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 3.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 3.0000 ± 0.0000
    pv_exact_mismatched: 3.0000 ± 0.0000
    pv_exact_match_rate: 50.0000 ± 0.0000
    average_pv_match_rate: 93.0315 ± 0.0000
    average_pv_mismatch_rate: 6.9685 ± 0.0000
    timing_matched: 5.0000 ± 0.0000
    timing_mismatched: 1.0000 ± 0.0000
    timing_match_rate: 83.3333 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 3.0000 ± 0.0000
    full_mismatched: 3.0000 ± 0.0000
    full_match_rate: 50.0000 ± 0.0000
    accuracy: 50.0000 ± 0.0000
    average_timing_score: 0.8479 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9138 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.6196 ± 0.0000
      comb_7_CodeBLEU: 0.7337 ± 0.0000
      dataflow_match_score: 0.6750 ± 0.0000
      ngram_match_score: 0.4207 ± 0.0000
      syntax_match_score: 0.7777 ± 0.0000
      weighted_ngram_match_score: 0.4384 ± 0.0000
    average_best_levenshtein: 40.0000 ± 0.0000
    average_best_normalized_levenshtein: 0.2631 ± 0.0000
    average_inference_time: 2.7852 ± 0.9797
==================================================
