Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 0.7353 ± 1.2736
total_entries: 136.0000 ± 0.0000
correct_matches: 1.0000 ± 1.7321
average_execution_time: 0.7805 ± 0.3720
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 1.0000 ± 1.7321
    exact_code_matches: 1.0000 ± 1.7321
    exact_code_match_rate: 0.8621 ± 1.4931
    pv_exact_matched: 1.0000 ± 1.7321
    pv_exact_mismatched: 115.0000 ± 1.7321
    pv_exact_match_rate: 0.8621 ± 1.4931
    average_pv_match_rate: 11.2069 ± 2.4637
    average_pv_mismatch_rate: 88.7931 ± 2.4637
    timing_matched: 27.0000 ± 8.6603
    timing_mismatched: 89.0000 ± 8.6603
    timing_match_rate: 23.2759 ± 7.4657
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 1.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 1.0000 ± 1.7321
    full_mismatched: 115.0000 ± 1.7321
    full_match_rate: 0.8621 ± 1.4931
    accuracy: 0.8621 ± 1.4931
    average_timing_score: 0.2328 ± 0.0747
    average_temp_score: 0.9914 ± 0.0000
    average_full_score: 0.1362 ± 0.0346
    average_best_codebleu:
      average_CodeBLEU: 0.2589 ± 0.0154
      comb_7_CodeBLEU: 0.4129 ± 0.0224
      dataflow_match_score: 0.0316 ± 0.0547
      ngram_match_score: 0.0023 ± 0.0041
      syntax_match_score: 0.0313 ± 0.0542
      weighted_ngram_match_score: 0.0019 ± 0.0033
    average_best_levenshtein: 187.4138 ± 288.9091
    average_best_normalized_levenshtein: 0.9842 ± 0.0274
    average_inference_time: 0.8176 ± 0.4361
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 0.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 0.0000 ± 0.0000
    pv_exact_mismatched: 20.0000 ± 0.0000
    pv_exact_match_rate: 0.0000 ± 0.0000
    average_pv_match_rate: 1.2500 ± 0.0000
    average_pv_mismatch_rate: 98.7500 ± 0.0000
    timing_matched: 1.0000 ± 0.0000
    timing_mismatched: 19.0000 ± 0.0000
    timing_match_rate: 5.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 9.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 0.0000 ± 0.0000
    full_mismatched: 20.0000 ± 0.0000
    full_match_rate: 0.0000 ± 0.0000
    accuracy: 0.0000 ± 0.0000
    average_timing_score: 0.0500 ± 0.0000
    average_temp_score: 0.5500 ± 0.0000
    average_full_score: 0.0200 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.2500 ± 0.0000
      comb_7_CodeBLEU: 0.4000 ± 0.0000
      dataflow_match_score: 0.0000 ± 0.0000
      ngram_match_score: 0.0000 ± 0.0000
      syntax_match_score: 0.0000 ± 0.0000
      weighted_ngram_match_score: 0.0000 ± 0.0000
    average_best_levenshtein: 256.0000 ± 0.0000
    average_best_normalized_levenshtein: 1.0000 ± 0.0000
    average_inference_time: 0.5647 ± 0.0006
==================================================
