Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 0.7937 ± 1.3746
total_entries: 126.0000 ± 0.0000
correct_matches: 1.0000 ± 1.7321
average_execution_time: 0.8708 ± 0.3864
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 1.0000 ± 1.7321
    exact_code_matches: 1.0000 ± 1.7321
    exact_code_match_rate: 0.8333 ± 1.4434
    pv_exact_matched: 1.0000 ± 1.7321
    pv_exact_mismatched: 119.0000 ± 1.7321
    pv_exact_match_rate: 0.8333 ± 1.4434
    average_pv_match_rate: 0.9722 ± 1.6839
    average_pv_mismatch_rate: 99.0278 ± 1.6839
    timing_matched: 1.3333 ± 2.3094
    timing_mismatched: 118.6667 ± 2.3094
    timing_match_rate: 1.1111 ± 1.9245
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 1.0000 ± 1.7321
    full_mismatched: 119.0000 ± 1.7321
    full_match_rate: 0.8333 ± 1.4434
    accuracy: 0.8333 ± 1.4434
    average_timing_score: 0.0111 ± 0.0192
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.0100 ± 0.0173
    average_best_codebleu:
      average_CodeBLEU: 0.2567 ± 0.0115
      comb_7_CodeBLEU: 0.4094 ± 0.0163
      dataflow_match_score: 0.0236 ± 0.0409
      ngram_match_score: 0.0023 ± 0.0039
      syntax_match_score: 0.0239 ± 0.0414
      weighted_ngram_match_score: 0.0018 ± 0.0032
    average_best_levenshtein: 150.2250 ± 224.5460
    average_best_normalized_levenshtein: 0.9856 ± 0.0250
    average_inference_time: 0.8694 ± 0.3840
  complex:
    count: 6.0000 ± 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: 6.0000 ± 0.0000
    pv_exact_match_rate: 0.0000 ± 0.0000
    average_pv_match_rate: 0.0000 ± 0.0000
    average_pv_mismatch_rate: 100.0000 ± 0.0000
    timing_matched: 0.0000 ± 0.0000
    timing_mismatched: 6.0000 ± 0.0000
    timing_match_rate: 0.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 0.0000 ± 0.0000
    full_mismatched: 6.0000 ± 0.0000
    full_match_rate: 0.0000 ± 0.0000
    accuracy: 0.0000 ± 0.0000
    average_timing_score: 0.0000 ± 0.0000
    average_temp_score: 0.1667 ± 0.0000
    average_full_score: 0.0000 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.2328 ± 0.0298
      comb_7_CodeBLEU: 0.3713 ± 0.0497
      dataflow_match_score: 0.0616 ± 0.1067
      ngram_match_score: 0.0006 ± 0.0010
      syntax_match_score: 0.0323 ± 0.0559
      weighted_ngram_match_score: 0.0033 ± 0.0058
    average_best_levenshtein: 181.5000 ± 107.0985
    average_best_normalized_levenshtein: 0.9550 ± 0.0780
    average_inference_time: 0.8992 ± 0.4349
==================================================
