Statistics Summary for op agent (mean ± std):
==================================================
accuracy: 87.7451 ± 0.4245
total_entries: 136.0000 ± 0.0000
correct_matches: 119.3333 ± 0.5774
average_execution_time: 1.6217 ± 0.1783
metrics_by_complexity:
  simple:
    count: 116.0000 ± 0.0000
    full_matches: 114.3333 ± 0.5774
    exact_code_matches: 103.3333 ± 0.5774
    exact_code_match_rate: 89.0805 ± 0.4977
    pv_exact_matched: 114.3333 ± 0.5774
    pv_exact_mismatched: 1.6667 ± 0.5774
    pv_exact_match_rate: 98.5632 ± 0.4977
    average_pv_match_rate: 99.5977 ± 0.0995
    average_pv_mismatch_rate: 0.4023 ± 0.0995
    timing_matched: 115.0000 ± 0.0000
    timing_mismatched: 1.0000 ± 0.0000
    timing_match_rate: 99.1379 ± 0.0000
    temp_matched: 0.0000 ± 0.0000
    temp_mismatched: 0.0000 ± 0.0000
    temp_match_rate: 0.0000 ± 0.0000
    full_matched: 114.3333 ± 0.5774
    full_mismatched: 1.6667 ± 0.5774
    full_match_rate: 98.5632 ± 0.4977
    accuracy: 98.5632 ± 0.4977
    average_timing_score: 0.9914 ± 0.0001
    average_temp_score: 1.0000 ± 0.0000
    average_full_score: 0.9951 ± 0.0008
    average_best_codebleu:
      average_CodeBLEU: 0.6111 ± 0.0002
      comb_7_CodeBLEU: 0.8429 ± 0.0008
      dataflow_match_score: 0.2816 ± 0.0050
      ngram_match_score: 0.2671 ± 0.0008
      syntax_match_score: 0.9948 ± 0.0023
      weighted_ngram_match_score: 0.1827 ± 0.0005
    average_best_levenshtein: 1.8937 ± 0.1294
    average_best_normalized_levenshtein: 0.0251 ± 0.0016
    average_inference_time: 0.8581 ± 0.1307
  complex:
    count: 20.0000 ± 0.0000
    full_matches: 5.0000 ± 0.0000
    exact_code_matches: 0.0000 ± 0.0000
    exact_code_match_rate: 0.0000 ± 0.0000
    pv_exact_matched: 5.0000 ± 0.0000
    pv_exact_mismatched: 15.0000 ± 0.0000
    pv_exact_match_rate: 25.0000 ± 0.0000
    average_pv_match_rate: 68.9789 ± 0.0000
    average_pv_mismatch_rate: 31.0211 ± 0.0000
    timing_matched: 12.0000 ± 0.0000
    timing_mismatched: 8.0000 ± 0.0000
    timing_match_rate: 60.0000 ± 0.0000
    temp_matched: 7.0000 ± 0.0000
    temp_mismatched: 2.0000 ± 0.0000
    temp_match_rate: 77.7778 ± 0.0000
    full_matched: 5.0000 ± 0.0000
    full_mismatched: 15.0000 ± 0.0000
    full_match_rate: 25.0000 ± 0.0000
    accuracy: 25.0000 ± 0.0000
    average_timing_score: 0.7496 ± 0.0027
    average_temp_score: 0.9131 ± 0.0000
    average_full_score: 0.7131 ± 0.0005
    average_best_codebleu:
      average_CodeBLEU: 0.3646 ± 0.0004
      comb_7_CodeBLEU: 0.4996 ± 0.0002
      dataflow_match_score: 0.5310 ± 0.0000
      ngram_match_score: 0.1296 ± 0.0008
      syntax_match_score: 0.6482 ± 0.0000
      weighted_ngram_match_score: 0.1498 ± 0.0008
    average_best_levenshtein: 198.0500 ± 0.3464
    average_best_normalized_levenshtein: 0.4933 ± 0.0015
    average_inference_time: 6.0508 ± 0.4571
==================================================
