Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 88.2353 ± 0.0000
total_entries: 136.0000 ± 0.0000
correct_matches: 120.0000 ± 0.0000
average_execution_time: 1.4275 ± 0.0946
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: 113.0000 ± 0.0000
    pv_exact_mismatched: 3.0000 ± 0.0000
    pv_exact_match_rate: 97.4138 ± 0.0000
    average_pv_match_rate: 98.6782 ± 0.0000
    average_pv_mismatch_rate: 1.3218 ± 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: 0.9798 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9854 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5963 ± 0.0000
      comb_7_CodeBLEU: 0.8347 ± 0.0000
      dataflow_match_score: 0.2328 ± 0.0000
      ngram_match_score: 0.2445 ± 0.0000
      syntax_match_score: 0.9871 ± 0.0000
      weighted_ngram_match_score: 0.1538 ± 0.0000
    average_best_levenshtein: 1.6810 ± 0.0000
    average_best_normalized_levenshtein: 0.0276 ± 0.0000
    average_inference_time: 1.0049 ± 0.0420
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 9.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 10.0000 ± 0.0000
    pv_exact_mismatched: 10.0000 ± 0.0000
    pv_exact_match_rate: 50.0000 ± 0.0000
    average_pv_match_rate: 70.3804 ± 0.0000
    average_pv_mismatch_rate: 29.6196 ± 0.0000
    timing_matched: 12.0000 ± 0.0000
    timing_mismatched: 8.0000 ± 0.0000
    timing_match_rate: 60.0000 ± 0.0000
    temp_matched: 4.0000 ± 0.0000
    temp_mismatched: 5.0000 ± 0.0000
    temp_match_rate: 44.4444 ± 0.0000
    full_matched: 9.0000 ± 0.0000
    full_mismatched: 11.0000 ± 0.0000
    full_match_rate: 45.0000 ± 0.0000
    accuracy: 45.0000 ± 0.0000
    average_timing_score: 0.7868 ± 0.0007
    average_temp_score: 0.8063 ± 0.0000
    average_full_score: 0.7195 ± 0.0001
    average_best_codebleu:
      average_CodeBLEU: 0.3836 ± 0.0000
      comb_7_CodeBLEU: 0.5131 ± 0.0000
      dataflow_match_score: 0.4824 ± 0.0000
      ngram_match_score: 0.1483 ± 0.0000
      syntax_match_score: 0.6166 ± 0.0000
      weighted_ngram_match_score: 0.1871 ± 0.0000
    average_best_levenshtein: 148.7500 ± 0.0000
    average_best_normalized_levenshtein: 0.4467 ± 0.0000
    average_inference_time: 3.8784 ± 0.3997
==================================================
