Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 84.1270 ± 0.0000
total_entries: 126.0000 ± 0.0000
correct_matches: 106.0000 ± 0.0000
average_execution_time: 0.7504 ± 0.1202
metrics_by_complexity:
  simple:
    count: 120.0000 ± 0.0000
    full_matches: 105.0000 ± 0.0000
    exact_code_matches: 100.0000 ± 0.0000
    exact_code_match_rate: 83.3333 ± 0.0000
    pv_exact_matched: 105.0000 ± 0.0000
    pv_exact_mismatched: 15.0000 ± 0.0000
    pv_exact_match_rate: 87.5000 ± 0.0000
    average_pv_match_rate: 93.7500 ± 0.0000
    average_pv_mismatch_rate: 6.2500 ± 0.0000
    timing_matched: 120.0000 ± 0.0000
    timing_mismatched: 0.0000 ± 0.0000
    timing_match_rate: 100.0000 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 105.0000 ± 0.0000
    full_mismatched: 15.0000 ± 0.0000
    full_match_rate: 87.5000 ± 0.0000
    accuracy: 87.5000 ± 0.0000
    average_timing_score: 1.0000 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9500 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.5753 ± 0.0000
      comb_7_CodeBLEU: 0.8228 ± 0.0000
      dataflow_match_score: 0.1000 ± 0.0000
      ngram_match_score: 0.2111 ± 0.0000
      syntax_match_score: 0.9756 ± 0.0000
      weighted_ngram_match_score: 0.1145 ± 0.0000
    average_best_levenshtein: 2.8083 ± 0.0000
    average_best_normalized_levenshtein: 0.0562 ± 0.0000
    average_inference_time: 0.6438 ± 0.0801
  complex:
    count: 6.0000 ± 0.0000
    full_matches: 1.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 2.0000 ± 0.0000
    pv_exact_mismatched: 4.0000 ± 0.0000
    pv_exact_match_rate: 33.3333 ± 0.0000
    average_pv_match_rate: 83.8359 ± 0.0000
    average_pv_mismatch_rate: 16.1641 ± 0.0000
    timing_matched: 4.0000 ± 0.0000
    timing_mismatched: 2.0000 ± 0.0000
    timing_match_rate: 66.6667 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 1.0000 ± 0.0000
    full_mismatched: 5.0000 ± 0.0000
    full_match_rate: 16.6667 ± 0.0000
    accuracy: 16.6667 ± 0.0000
    average_timing_score: 0.8024 ± 0.0000
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.8312 ± 0.0000
    average_best_codebleu:
      average_CodeBLEU: 0.6373 ± 0.0000
      comb_7_CodeBLEU: 0.7091 ± 0.0000
      dataflow_match_score: 0.7391 ± 0.0000
      ngram_match_score: 0.4904 ± 0.0000
      syntax_match_score: 0.7749 ± 0.0000
      weighted_ngram_match_score: 0.5447 ± 0.0000
    average_best_levenshtein: 33.5000 ± 0.0000
    average_best_normalized_levenshtein: 0.1964 ± 0.0000
    average_inference_time: 2.8823 ± 0.9238
==================================================
