{
  "description": "Controlled data-quality experiments; not a comparison with frontier models or a before/after validation test.",
  "source_revision": "1fcefb66637638ead2c50ab60d49ffaada57d769",
  "source_repository": "Thatayotlhe04/a-parallel-paradigm",
  "author": "Thatayotlhe Tsenang",
  "date": "2026-11-08",
  "accuracy": {
    "baselines": {
      "clean_exact_match": [
        1.0,
        1.0,
        1.0,
        5
      ],
      "clean_exact_match_sampled": [
        0.9826666666666666,
        0.9754088600155582,
        0.989924473317775,
        5
      ],
      "clean_empty_rate_sampled": [
        0.0,
        0.0,
        0.0,
        5
      ],
      "clean_gold_nll": [
        0.0015108491992577911,
        0.0014311601529985436,
        0.0015905382455170386,
        5
      ],
      "lowdata_clean_exact_match": [
        0.1,
        0.04420181864810306,
        0.15579818135189694,
        5
      ],
      "lowdata_clean_exact_match_sampled": [
        0.07166666666666667,
        0.028679062751446834,
        0.11465427058188651,
        5
      ],
      "lowdata_clean_empty_rate_sampled": [
        0.0,
        0.0,
        0.0,
        5
      ],
      "lowdata_clean_gold_nll": [
        0.7823761463165283,
        0.6250337275937919,
        0.9397185650392647,
        5
      ]
    },
    "conditions": [
      {
        "condition": "contamination",
        "rate": 0.1,
        "exact_match_mean": 0.9996666666666666,
        "exact_match_diff": -0.000333333333333341,
        "exact_match_lo": -0.0012588150350659603,
        "exact_match_hi": 0.0005921483683992781,
        "n_seeds": 5,
        "exact_match_p": 0.37390096630005887,
        "exact_match_sampled_mean": 0.985,
        "exact_match_sampled_diff": 0.0023333333333333426,
        "exact_match_sampled_lo": -0.0038056177766398478,
        "exact_match_sampled_hi": 0.008472284443306534,
        "exact_match_sampled_p": 0.35081481481480736,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.001551277027465403,
        "gold_nll_diff": 4.042782820761204e-05,
        "gold_nll_lo": -5.654838305662215e-05,
        "gold_nll_hi": 0.00013740403947184624,
        "gold_nll_p": 0.31149497509042623,
        "inflation_mean": 3.7037037037035424e-05,
        "inflation_lo": -6.579426315547098e-05,
        "inflation_hi": 0.00013986833722954181,
        "acc_seen_mean": 1.0,
        "acc_unseen_mean": 0.9996296296296296
      },
      {
        "condition": "contamination",
        "rate": 0.3,
        "exact_match_mean": 1.0,
        "exact_match_diff": 0.0,
        "exact_match_lo": 0.0,
        "exact_match_hi": 0.0,
        "n_seeds": 5,
        "exact_match_p": 1.0,
        "exact_match_sampled_mean": 0.9869999999999999,
        "exact_match_sampled_diff": 0.004333333333333323,
        "exact_match_sampled_lo": -0.0008609198350844841,
        "exact_match_sampled_hi": 0.00952758650175113,
        "exact_match_sampled_p": 0.08146997954982481,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.0014879752648994326,
        "gold_nll_diff": -2.2873934358358384e-05,
        "gold_nll_lo": -7.866336611608487e-05,
        "gold_nll_hi": 3.291549739936809e-05,
        "gold_nll_p": 0.31852463990176494,
        "inflation_mean": 0.0,
        "inflation_lo": 0.0,
        "inflation_hi": 0.0,
        "acc_seen_mean": 1.0,
        "acc_unseen_mean": 1.0
      },
      {
        "condition": "dup_concentrated",
        "rate": 0.1,
        "exact_match_mean": 1.0,
        "exact_match_diff": 0.0,
        "exact_match_lo": 0.0,
        "exact_match_hi": 0.0,
        "n_seeds": 5,
        "exact_match_p": 1.0,
        "exact_match_sampled_mean": 0.9793333333333333,
        "exact_match_sampled_diff": -0.0033333333333333435,
        "exact_match_sampled_lo": -0.009186593553900491,
        "exact_match_sampled_hi": 0.0025199268872338034,
        "exact_match_sampled_p": 0.18900365845517003,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.0015205215429887175,
        "gold_nll_diff": 9.672343730926514e-06,
        "gold_nll_lo": -8.007319453145817e-05,
        "gold_nll_hi": 9.94178819933112e-05,
        "gold_nll_p": 0.7796663824105573
      },
      {
        "condition": "dup_concentrated",
        "rate": 0.3,
        "exact_match_mean": 1.0,
        "exact_match_diff": 0.0,
        "exact_match_lo": 0.0,
        "exact_match_hi": 0.0,
        "n_seeds": 5,
        "exact_match_p": 1.0,
        "exact_match_sampled_mean": 0.9843333333333334,
        "exact_match_sampled_diff": 0.001666666666666683,
        "exact_match_sampled_lo": -0.002723278498758718,
        "exact_match_sampled_hi": 0.006056611832092084,
        "exact_match_sampled_p": 0.35130100296325456,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.0017274056794121862,
        "gold_nll_diff": 0.0002165564801543951,
        "gold_nll_lo": 0.00013427969659451482,
        "gold_nll_hi": 0.0002988332637142754,
        "gold_nll_p": 0.001864914504957384
      },
      {
        "condition": "empty_target",
        "rate": 0.1,
        "exact_match_mean": 0.954,
        "exact_match_diff": -0.046,
        "exact_match_lo": -0.059379533753855164,
        "exact_match_hi": -0.032620466246144834,
        "n_seeds": 5,
        "exact_match_p": 0.0006726708053639232,
        "exact_match_sampled_mean": 0.8996666666666666,
        "exact_match_sampled_diff": -0.08300000000000003,
        "exact_match_sampled_lo": -0.09721751967598709,
        "exact_match_sampled_hi": -0.06878248032401298,
        "exact_match_sampled_p": 8.476895894487632e-05,
        "empty_rate_sampled_mean": 0.07833333333333334,
        "empty_rate_sampled_diff": 0.07833333333333334,
        "empty_rate_sampled_lo": 0.06362719903464455,
        "empty_rate_sampled_hi": 0.09303946763202213,
        "empty_rate_sampled_p": 0.00012169684452001326,
        "gold_nll_mean": 0.014243058860301971,
        "gold_nll_diff": 0.01273220966104418,
        "gold_nll_lo": 0.009413307600520542,
        "gold_nll_hi": 0.01605111172156782,
        "gold_nll_p": 0.0004400054660178182
      },
      {
        "condition": "empty_target",
        "rate": 0.3,
        "exact_match_mean": 0.776,
        "exact_match_diff": -0.22399999999999998,
        "exact_match_lo": -0.26004035364194955,
        "exact_match_hi": -0.1879596463580504,
        "n_seeds": 5,
        "exact_match_p": 6.617563236548885e-05,
        "exact_match_sampled_mean": 0.6886666666666666,
        "exact_match_sampled_diff": -0.29400000000000004,
        "exact_match_sampled_lo": -0.3168389683560204,
        "exact_match_sampled_hi": -0.2711610316439797,
        "exact_match_sampled_p": 3.658038719254428e-06,
        "empty_rate_sampled_mean": 0.284,
        "empty_rate_sampled_diff": 0.284,
        "empty_rate_sampled_lo": 0.2645538697454536,
        "empty_rate_sampled_hi": 0.30344613025454636,
        "empty_rate_sampled_p": 2.2105248253114557e-06,
        "gold_nll_mean": 0.05256788358092308,
        "gold_nll_diff": 0.05105703438166529,
        "gold_nll_lo": 0.03571747853599316,
        "gold_nll_hi": 0.06639659022733742,
        "gold_nll_p": 0.00076218380780538
      },
      {
        "condition": "label_noise_random",
        "rate": 0.1,
        "exact_match_mean": 0.9886666666666667,
        "exact_match_diff": -0.011333333333333329,
        "exact_match_lo": -0.017437304380161892,
        "exact_match_hi": -0.0052293622865047665,
        "n_seeds": 5,
        "exact_match_p": 0.006720880690790092,
        "exact_match_sampled_mean": 0.908,
        "exact_match_sampled_diff": -0.07466666666666669,
        "exact_match_sampled_lo": -0.08762341049092304,
        "exact_match_sampled_hi": -0.06170992284241033,
        "exact_match_sampled_p": 8.92165423810604e-05,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.010199067182838917,
        "gold_nll_diff": 0.008688217983581126,
        "gold_nll_lo": 0.007877446334010921,
        "gold_nll_hi": 0.009498989633151331,
        "gold_nll_p": 7.599806851539109e-06
      },
      {
        "condition": "label_noise_random",
        "rate": 0.3,
        "exact_match_mean": 0.9630000000000001,
        "exact_match_diff": -0.03699999999999999,
        "exact_match_lo": -0.04827797333149523,
        "exact_match_hi": -0.025722026668504754,
        "n_seeds": 5,
        "exact_match_p": 0.0008057456843671433,
        "exact_match_sampled_mean": 0.7546666666666667,
        "exact_match_sampled_diff": -0.22800000000000004,
        "exact_match_sampled_lo": -0.24515264925931504,
        "exact_match_sampled_hi": -0.21084735074068503,
        "exact_match_sampled_p": 3.218530169492178e-06,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.03178755827248096,
        "gold_nll_diff": 0.030276709073223173,
        "gold_nll_lo": 0.028179738348850715,
        "gold_nll_hi": 0.03237367979759563,
        "gold_nll_p": 2.313822981690895e-06
      },
      {
        "condition": "label_noise_systematic",
        "rate": 0.1,
        "exact_match_mean": 0.8780000000000001,
        "exact_match_diff": -0.12199999999999997,
        "exact_match_lo": -0.13838651534664476,
        "exact_match_hi": -0.10561348465335518,
        "n_seeds": 5,
        "exact_match_p": 3.235610872143166e-05,
        "exact_match_sampled_mean": 0.8646666666666667,
        "exact_match_sampled_diff": -0.11800000000000002,
        "exact_match_sampled_lo": -0.13540053276240419,
        "exact_match_sampled_hi": -0.10059946723759586,
        "exact_match_sampled_p": 4.685906526376709e-05,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.13127212077379227,
        "gold_nll_diff": 0.1297612715745345,
        "gold_nll_lo": 0.10067908648709119,
        "gold_nll_hi": 0.15884345666197777,
        "gold_nll_p": 0.00024405537481082666,
        "acc_on_affected_mean": 0.0,
        "rows_affected_share_mean": 0.1214
      },
      {
        "condition": "label_noise_systematic",
        "rate": 0.3,
        "exact_match_mean": 0.6833333333333333,
        "exact_match_diff": -0.3166666666666667,
        "exact_match_lo": -0.33070229160664105,
        "exact_match_hi": -0.30263104172669236,
        "n_seeds": 5,
        "exact_match_p": 3.890227521280751e-07,
        "exact_match_sampled_mean": 0.673,
        "exact_match_sampled_diff": -0.30966666666666665,
        "exact_match_sampled_lo": -0.3216801178806498,
        "exact_match_sampled_hi": -0.2976532154526835,
        "exact_match_sampled_p": 2.2841422074289615e-07,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.3342846155166626,
        "gold_nll_diff": 0.3327737663174048,
        "gold_nll_lo": 0.26429083467448483,
        "gold_nll_hi": 0.40125669796032476,
        "gold_nll_p": 0.00017465712824587116,
        "acc_on_affected_mean": 0.0,
        "rows_affected_share_mean": 0.3194
      },
      {
        "condition": "lowdata_contamination",
        "rate": 0.1,
        "exact_match_mean": 0.19966666666666666,
        "exact_match_diff": 0.09966666666666665,
        "exact_match_lo": 0.027585959382767505,
        "exact_match_hi": 0.1717473739505658,
        "n_seeds": 5,
        "exact_match_p": 0.018475062571887985,
        "exact_match_sampled_mean": 0.17266666666666666,
        "exact_match_sampled_diff": 0.101,
        "exact_match_sampled_lo": 0.031780402839419866,
        "exact_match_sampled_hi": 0.17021959716058016,
        "exact_match_sampled_p": 0.015460687335405648,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.6987105548381806,
        "gold_nll_diff": -0.08366559147834778,
        "gold_nll_lo": -0.23433812586326486,
        "gold_nll_hi": 0.06700694290656932,
        "gold_nll_p": 0.1980029483892812,
        "inflation_mean": 0.08892592592592592,
        "inflation_lo": 0.07877032629394702,
        "inflation_hi": 0.09908152555790482,
        "acc_seen_mean": 1.0,
        "acc_unseen_mean": 0.11074074074074074
      },
      {
        "condition": "lowdata_contamination",
        "rate": 0.3,
        "exact_match_mean": 0.3496666666666667,
        "exact_match_diff": 0.24966666666666665,
        "exact_match_lo": 0.19553587564337363,
        "exact_match_hi": 0.3037974576899597,
        "n_seeds": 5,
        "exact_match_p": 0.00021432737763210232,
        "exact_match_sampled_mean": 0.33199999999999996,
        "exact_match_sampled_diff": 0.26033333333333336,
        "exact_match_sampled_lo": 0.21836909048567463,
        "exact_match_sampled_hi": 0.30229757618099207,
        "exact_match_sampled_p": 6.66648087091266e-05,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.6090388417243957,
        "gold_nll_diff": -0.17333730459213256,
        "gold_nll_lo": -0.34431382713467795,
        "gold_nll_hi": -0.0023607820495871457,
        "gold_nll_p": 0.04808250317284293,
        "inflation_mean": 0.2787142857142857,
        "inflation_lo": 0.26881031653704063,
        "inflation_hi": 0.28861825489153076,
        "acc_seen_mean": 1.0,
        "acc_unseen_mean": 0.07095238095238096
      },
      {
        "condition": "truncated_target",
        "rate": 0.1,
        "exact_match_mean": 1.0,
        "exact_match_diff": 0.0,
        "exact_match_lo": 0.0,
        "exact_match_hi": 0.0,
        "n_seeds": 5,
        "exact_match_p": 1.0,
        "exact_match_sampled_mean": 0.985,
        "exact_match_sampled_diff": 0.0023333333333333205,
        "exact_match_sampled_lo": -0.0032579856415444167,
        "exact_match_sampled_hi": 0.007924652308211058,
        "exact_match_sampled_p": 0.3110607673343719,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.0015635509276762604,
        "gold_nll_diff": 5.270172841846943e-05,
        "gold_nll_lo": 2.3081422385644075e-05,
        "gold_nll_hi": 8.232203445129479e-05,
        "gold_nll_p": 0.007816943000957694
      },
      {
        "condition": "truncated_target",
        "rate": 0.3,
        "exact_match_mean": 0.9996666666666666,
        "exact_match_diff": -0.000333333333333341,
        "exact_match_lo": -0.0012588150350659603,
        "exact_match_hi": 0.0005921483683992781,
        "n_seeds": 5,
        "exact_match_p": 0.37390096630005887,
        "exact_match_sampled_mean": 0.9823333333333334,
        "exact_match_sampled_diff": -0.000333333333333341,
        "exact_match_sampled_lo": -0.006075796275252183,
        "exact_match_sampled_hi": 0.005409129608585501,
        "exact_match_sampled_p": 0.8797762036739447,
        "empty_rate_sampled_mean": 0.0,
        "empty_rate_sampled_diff": 0.0,
        "empty_rate_sampled_lo": 0.0,
        "empty_rate_sampled_hi": 0.0,
        "empty_rate_sampled_p": 1.0,
        "gold_nll_mean": 0.0018418311607092618,
        "gold_nll_diff": 0.00033098196145147084,
        "gold_nll_lo": -3.210392072424458e-06,
        "gold_nll_hi": 0.0006651743149753661,
        "gold_nll_p": 0.05138639509539451
      }
    ]
  },
  "detection": {
    "meta": {
      "alpaca_sha256": "bd844b8247a0f543804b6ce0882b0aaec4bbf5e8d66167df6213a0f1e4fe878b",
      "hh_sha256": "518a5bf288456fc9f3b7c980c54116fba0c52f274d3f4d344675d83e4058f6f4",
      "alpaca_rows": 51760,
      "hh_pairs_parsed": 43709,
      "seed": 20261108,
      "natural_full_alpaca": {
        "seconds": 6.39,
        "rows_per_sec": 8104,
        "rows_with_error": 4,
        "rows_with_any": 61,
        "by_rule": {
          "duplicates": 4,
          "pii": 57
        }
      },
      "base_raw_flags": {
        "parallelogram": 13,
        "cookbook": 0,
        "exact": 0,
        "minhash": 0,
        "ngram13": 24
      },
      "base_clean_rows": 9963,
      "base_clean_flags": {
        "parallelogram": 0,
        "cookbook": 0,
        "exact": 0,
        "minhash": 0,
        "ngram13": 0
      },
      "pref_raw_flags": {
        "parallelogram": 28,
        "exact": 0,
        "minhash": 12,
        "parallelogram_by_rule": {
          "preference": 7,
          "empty-content": 9,
          "roles": 5,
          "pii": 6,
          "encoding": 1
        }
      },
      "pref_clean_rows": 4960,
      "fix": {
        "summary": {
          "total": 9963,
          "unchanged": 9813,
          "fixed": 50,
          "dropped": 90,
          "unparseable": 10,
          "emitted": 9863
        },
        "clean_rows": 9773,
        "clean_rows_kept_unchanged": 9768,
        "injected_fate": {
          "invalid_json": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "missing_assistant": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "role_order": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "empty_response": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "whitespace_response": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "overlong_response": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "overlong_user": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "secret_aws": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "secret_github": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "pii_email": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "mojibake": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "label_swap": {
            "kept_unchanged": 10,
            "removed_or_changed": 0
          },
          "truncated_response": {
            "kept_unchanged": 10,
            "removed_or_changed": 0
          },
          "exact_duplicate": {
            "kept_unchanged": 10,
            "removed_or_changed": 0
          },
          "whitespace_duplicate": {
            "kept_unchanged": 5,
            "removed_or_changed": 5
          },
          "near_duplicate": {
            "kept_unchanged": 10,
            "removed_or_changed": 0
          },
          "eval_exact": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "eval_prompt_only": {
            "kept_unchanged": 0,
            "removed_or_changed": 10
          },
          "eval_paraphrase": {
            "kept_unchanged": 10,
            "removed_or_changed": 0
          }
        }
      },
      "cookbook_seconds_10k": 0.086,
      "minhash_seconds_10k": 9.68
    },
    "results": [
      {
        "file": "sft",
        "defect": "invalid_json",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 1.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "schema"
        ],
        "cookbook_crashed": true,
        "P_seconds": 1.36
      },
      {
        "file": "sft",
        "defect": "missing_assistant",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 1.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "roles"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.36
      },
      {
        "file": "sft",
        "defect": "role_order",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "roles"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.42
      },
      {
        "file": "sft",
        "defect": "empty_response",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 1.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "empty-content"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.49
      },
      {
        "file": "sft",
        "defect": "whitespace_response",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "empty-content"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.41
      },
      {
        "file": "sft",
        "defect": "overlong_response",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 0.98,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "context-window"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.64
      },
      {
        "file": "sft",
        "defect": "overlong_user",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "context-window"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.67
      },
      {
        "file": "sft",
        "defect": "secret_aws",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "secrets"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.51
      },
      {
        "file": "sft",
        "defect": "secret_github",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "secrets"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.44
      },
      {
        "file": "sft",
        "defect": "pii_email",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "pii"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.57
      },
      {
        "file": "sft",
        "defect": "mojibake",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "encoding"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.45
      },
      {
        "file": "sft",
        "defect": "label_swap",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 0.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [],
        "cookbook_crashed": false,
        "P_seconds": 1.43
      },
      {
        "file": "sft",
        "defect": "truncated_response",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 0.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [],
        "cookbook_crashed": false,
        "P_seconds": 1.53
      },
      {
        "file": "sft",
        "defect": "exact_duplicate",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 1.0,
        "exact_collateral": 0,
        "minhash_recall": 1.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "duplicates"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.5
      },
      {
        "file": "sft",
        "defect": "whitespace_duplicate",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 1.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "duplicates"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.46
      },
      {
        "file": "sft",
        "defect": "near_duplicate",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 0.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.16,
        "minhash_collateral": 0,
        "ngram13_recall": 0.0,
        "ngram13_collateral": 0,
        "P_rules": [],
        "cookbook_crashed": false,
        "P_seconds": 1.6
      },
      {
        "file": "sft",
        "defect": "eval_exact",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 1.0,
        "ngram13_collateral": 0,
        "P_rules": [
          "eval-overlap"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.47
      },
      {
        "file": "sft",
        "defect": "eval_prompt_only",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.46,
        "ngram13_collateral": 0,
        "P_rules": [
          "eval-overlap"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.42
      },
      {
        "file": "sft",
        "defect": "eval_paraphrase",
        "n_injected": 100,
        "rows": 9963,
        "P_error_recall": 0.02,
        "P_error_collateral": 0,
        "P_any_recall": 0.02,
        "P_any_collateral": 0,
        "cookbook_recall": 0.0,
        "cookbook_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "ngram13_recall": 0.96,
        "ngram13_collateral": 0,
        "P_rules": [
          "eval-overlap"
        ],
        "cookbook_crashed": false,
        "P_seconds": 1.42
      },
      {
        "file": "pref",
        "defect": "identical_pair",
        "n_injected": 100,
        "rows": 4960,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "P_rules": [
          "preference"
        ]
      },
      {
        "file": "pref",
        "defect": "empty_chosen",
        "n_injected": 100,
        "rows": 4960,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.01,
        "minhash_collateral": 0,
        "P_rules": [
          "empty-content"
        ]
      },
      {
        "file": "pref",
        "defect": "flipped_pair",
        "n_injected": 100,
        "rows": 4960,
        "P_error_recall": 0.0,
        "P_error_collateral": 0,
        "P_any_recall": 0.0,
        "P_any_collateral": 0,
        "exact_recall": 0.0,
        "exact_collateral": 0,
        "minhash_recall": 0.0,
        "minhash_collateral": 0,
        "P_rules": []
      },
      {
        "file": "pref",
        "defect": "duplicate_pair",
        "n_injected": 100,
        "rows": 4960,
        "P_error_recall": 1.0,
        "P_error_collateral": 0,
        "P_any_recall": 1.0,
        "P_any_collateral": 0,
        "exact_recall": 1.0,
        "exact_collateral": 0,
        "minhash_recall": 1.0,
        "minhash_collateral": 0,
        "P_rules": [
          "duplicates"
        ]
      }
    ]
  }
}
