Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 40.4412 ± 0.0000
total_entries: 136.0000 ± 0.0000
correct_matches: 55.0000 ± 0.0000
average_execution_time: 0.9301 ± 0.0435
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 55.0000 ± 0.0000
    exact_code_matches: 28.0000 ± 0.0000
    exact_code_match_rate: 24.1379 ± 0.0000
    pv_exact_matched: 55.0000 ± 0.0000
    pv_exact_mismatched: 61.0000 ± 0.0000
    pv_exact_match_rate: 47.4138 ± 0.0000
    average_pv_match_rate: 53.8793 ± 0.0000
    average_pv_mismatch_rate: 46.1207 ± 0.0000
    timing_matched: 71.6667 ± 0.5774
    timing_mismatched: 44.3333 ± 0.5774
    timing_match_rate: 61.7816 ± 0.4977
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 55.0000 ± 0.0000
    full_mismatched: 61.0000 ± 0.0000
    full_match_rate: 47.4138 ± 0.0000
    accuracy: 47.4138 ± 0.0000
    average_timing_score: 0.6166 ± 0.0013
    average_temp_score: 0.9052 ± 0.0000
    average_full_score: 0.5544 ± 0.0003
    average_best_codebleu:
      average_CodeBLEU: 0.4895 ± 0.0000
      comb_7_CodeBLEU: 0.7403 ± 0.0000
      dataflow_match_score: 0.2241 ± 0.0000
      ngram_match_score: 0.0793 ± 0.0000
      syntax_match_score: 0.8151 ± 0.0000
      weighted_ngram_match_score: 0.0637 ± 0.0000
    average_best_levenshtein: 71.2644 ± 0.0796
    average_best_normalized_levenshtein: 0.3585 ± 0.0004
    average_inference_time: 0.8289 ± 0.0440
  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: 9.9056 ± 0.1997
    average_pv_mismatch_rate: 90.0944 ± 0.1997
    timing_matched: 5.6667 ± 0.5774
    timing_mismatched: 14.3333 ± 0.5774
    timing_match_rate: 28.3333 ± 2.8868
    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: 0.3337 ± 0.0339
    average_temp_score: 0.5500 ± 0.0000
    average_full_score: 0.1377 ± 0.0073
    average_best_codebleu:
      average_CodeBLEU: 0.2676 ± 0.0005
      comb_7_CodeBLEU: 0.3834 ± 0.0006
      dataflow_match_score: 0.3898 ± 0.0014
      ngram_match_score: 0.0690 ± 0.0003
      syntax_match_score: 0.4315 ± 0.0000
      weighted_ngram_match_score: 0.0800 ± 0.0003
    average_best_levenshtein: 211.0833 ± 6.7839
    average_best_normalized_levenshtein: 0.6469 ± 0.0027
    average_inference_time: 1.5167 ± 0.0456
==================================================
