Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 86.7647 ± 0.0000
total_entries: 136.0000 ± 0.0000
correct_matches: 118.0000 ± 0.0000
average_execution_time: 3.5212 ± 0.1931
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 111.0000 ± 0.0000
    exact_code_matches: 104.0000 ± 0.0000
    exact_code_match_rate: 89.6552 ± 0.0000
    pv_exact_matched: 114.0000 ± 0.0000
    pv_exact_mismatched: 2.0000 ± 0.0000
    pv_exact_match_rate: 98.2759 ± 0.0000
    average_pv_match_rate: 99.2816 ± 0.0000
    average_pv_mismatch_rate: 0.7184 ± 0.0000
    timing_matched: 112.0000 ± 0.0000
    timing_mismatched: 4.0000 ± 0.0000
    timing_match_rate: 96.5517 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 111.0000 ± 0.0000
    full_mismatched: 5.0000 ± 0.0000
    full_match_rate: 95.6897 ± 0.0000
    accuracy: 95.6897 ± 0.0000
    average_timing_score: 0.9833 ± 0.0004
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9909 ± 0.0001
    average_best_codebleu:
      average_CodeBLEU: 0.6287 ± 0.0000
      comb_7_CodeBLEU: 0.8495 ± 0.0000
      dataflow_match_score: 0.3103 ± 0.0000
      ngram_match_score: 0.3040 ± 0.0000
      syntax_match_score: 0.9933 ± 0.0000
      weighted_ngram_match_score: 0.2176 ± 0.0000
    average_best_levenshtein: 1.5862 ± 0.0000
    average_best_normalized_levenshtein: 0.0230 ± 0.0000
    average_inference_time: 2.8836 ± 0.1458
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 7.0000 ± 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: 83.3532 ± 0.4535
    average_pv_mismatch_rate: 16.6468 ± 0.4535
    timing_matched: 14.0000 ± 0.0000
    timing_mismatched: 6.0000 ± 0.0000
    timing_match_rate: 70.0000 ± 0.0000
    temp_matched: 8.0000 ± 0.0000
    temp_mismatched: 1.0000 ± 0.0000
    temp_match_rate: 88.8889 ± 0.0000
    full_matched: 7.0000 ± 0.0000
    full_mismatched: 13.0000 ± 0.0000
    full_match_rate: 35.0000 ± 0.0000
    accuracy: 35.0000 ± 0.0000
    average_timing_score: 0.7838 ± 0.0048
    average_temp_score: 0.9859 ± 0.0000
    average_full_score: 0.8276 ± 0.0042
    average_best_codebleu:
      average_CodeBLEU: 0.4201 ± 0.0022
      comb_7_CodeBLEU: 0.5447 ± 0.0023
      dataflow_match_score: 0.4951 ± 0.0028
      ngram_match_score: 0.1864 ± 0.0020
      syntax_match_score: 0.6606 ± 0.0024
      weighted_ngram_match_score: 0.2383 ± 0.0022
    average_best_levenshtein: 166.7667 ± 0.4368
    average_best_normalized_levenshtein: 0.4684 ± 0.0016
    average_inference_time: 7.2197 ± 0.5859
==================================================
