docs(semantics): record iterative retraining study

Co-authored-by: Rocky <hkgood@users.noreply.github.com>
This commit is contained in:
Cursor Agent
2026-08-27 10:23:44 +00:00
parent 00b90d9d28
commit de9cd332a5
13 changed files with 71684 additions and 72 deletions
@@ -0,0 +1,629 @@
{
"automaticPromotionAllowed": false,
"baseCorpus": "ModelTraining/ClipboardSemantics/clipboard_semantic_corpus.jsonl",
"configuration": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"configurationsByIntent": {
"blessing": {
"alpha": 1.2e-06,
"augmentation": "numeric",
"char_max": 8,
"char_min": 3,
"external_weight": 0.93,
"hard_example_weight": 2.7,
"l1_ratio": 0.25,
"max_char_features": 95000,
"max_word_features": 45000,
"min_df": 1,
"pseudo_weight": 0.43499999999999994,
"round": 19,
"word_max": 3
},
"complaint": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"confirmationDecision": {
"alpha": 5e-06,
"augmentation": "prefix",
"char_max": 6,
"char_min": 2,
"external_weight": 0.77,
"hard_example_weight": 1.8,
"l1_ratio": 0.1,
"max_char_features": 70000,
"max_word_features": 30000,
"min_df": 2,
"pseudo_weight": 0.31499999999999995,
"round": 11,
"word_max": 2
},
"followUpReminder": {
"alpha": 2.5e-06,
"augmentation": "prefix",
"char_max": 7,
"char_min": 3,
"external_weight": 0.8300000000000001,
"hard_example_weight": 2.2,
"l1_ratio": 0.2,
"max_char_features": 80000,
"max_word_features": 40000,
"min_df": 1,
"pseudo_weight": 0.36,
"round": 14,
"word_max": 3
},
"invitation": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"question": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"replyableMessage": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"scheduleNegotiation": {
"alpha": 6e-06,
"augmentation": "punctuation",
"char_max": 6,
"char_min": 3,
"external_weight": 0.6900000000000001,
"hard_example_weight": 1.5,
"l1_ratio": 0.1,
"max_char_features": 65000,
"max_word_features": 32000,
"min_df": 2,
"pseudo_weight": 0.255,
"round": 7,
"word_max": 3
},
"task": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
}
},
"openCorpus": "ModelTraining/ClipboardSemantics/open-training-corpus.jsonl",
"reason": "Surrogate feature weights are not deployable NLModel assets and cannot bypass the existing macOS acceptance policy.",
"requiredCommands": [
"python3 Scripts/clipboard_semantics/generate_open_training_corpus.py",
"xcrun swift Scripts/clipboard_semantics/train_models.swift --algorithms maxEnt --corpus ModelTraining/ClipboardSemantics/combined-training-corpus.jsonl",
"xcrun swift Scripts/clipboard_semantics/evaluate_random_holdout.swift --corpus ModelTraining/ClipboardSemantics/random-holdout-corpus.jsonl"
],
"schemaVersion": 1,
"seed": 20260827,
"selectedRound": 1,
"selectedRoundsByIntent": {
"blessing": 19,
"complaint": 1,
"confirmationDecision": 11,
"followUpReminder": 14,
"invitation": 1,
"question": 1,
"replyableMessage": 1,
"scheduleNegotiation": 7,
"task": 1
},
"thresholds": {
"blessing": {
"abstained": true,
"byLanguage": {
"en": {
"abstained": true,
"metrics": {
"f1": 0.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 0.0,
"predictedPositive": 0,
"recall": 0.0,
"total": 0,
"trueNegative": 0,
"truePositive": 0,
"wilsonPrecisionLower95": 0
},
"threshold": 1.0
},
"zh-Hans": {
"abstained": true,
"metrics": {
"f1": 0.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 0.0,
"predictedPositive": 0,
"recall": 0.0,
"total": 0,
"trueNegative": 0,
"truePositive": 0,
"wilsonPrecisionLower95": 0
},
"threshold": 1.0
}
},
"metrics": {
"f1": 0.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 0.0,
"predictedPositive": 0,
"recall": 0.0,
"total": 0,
"trueNegative": 0,
"truePositive": 0,
"wilsonPrecisionLower95": 0
},
"threshold": 1.0
},
"complaint": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 180,
"recall": 1.0,
"total": 1485,
"trueNegative": 1305,
"truePositive": 180,
"wilsonPrecisionLower95": 0.9791
},
"threshold": 0.5
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.9526,
"falseNegative": 9,
"falsePositive": 8,
"precision": 0.9553,
"predictedPositive": 179,
"recall": 0.95,
"total": 1485,
"trueNegative": 1297,
"truePositive": 171,
"wilsonPrecisionLower95": 0.9143
},
"threshold": 0.54
}
},
"metrics": {
"f1": 0.9764,
"falseNegative": 9,
"falsePositive": 8,
"precision": 0.9777,
"predictedPositive": 359,
"recall": 0.975,
"total": 2970,
"trueNegative": 2602,
"truePositive": 351,
"wilsonPrecisionLower95": 0.9567
},
"threshold": 0.54
},
"confirmationDecision": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.931,
"falseNegative": 9,
"falsePositive": 3,
"precision": 0.9643,
"predictedPositive": 84,
"recall": 0.9,
"total": 1485,
"trueNegative": 1392,
"truePositive": 81,
"wilsonPrecisionLower95": 0.9002
},
"threshold": 0.97
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.7222,
"falseNegative": 38,
"falsePositive": 2,
"precision": 0.963,
"predictedPositive": 54,
"recall": 0.5778,
"total": 1485,
"trueNegative": 1393,
"truePositive": 52,
"wilsonPrecisionLower95": 0.8746
},
"threshold": 0.86
}
},
"metrics": {
"f1": 0.8077,
"falseNegative": 54,
"falsePositive": 6,
"precision": 0.9545,
"predictedPositive": 132,
"recall": 0.7,
"total": 2970,
"trueNegative": 2784,
"truePositive": 126,
"wilsonPrecisionLower95": 0.9044
},
"threshold": 0.95
},
"followUpReminder": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8788,
"falseNegative": 35,
"falsePositive": 5,
"precision": 0.9667,
"predictedPositive": 150,
"recall": 0.8056,
"total": 1485,
"trueNegative": 1300,
"truePositive": 145,
"wilsonPrecisionLower95": 0.9243
},
"threshold": 0.845
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.6113,
"falseNegative": 99,
"falsePositive": 4,
"precision": 0.9529,
"predictedPositive": 85,
"recall": 0.45,
"total": 1485,
"trueNegative": 1301,
"truePositive": 81,
"wilsonPrecisionLower95": 0.8852
},
"threshold": 0.97
}
},
"metrics": {
"f1": 0.7682,
"falseNegative": 128,
"falsePositive": 12,
"precision": 0.9508,
"predictedPositive": 244,
"recall": 0.6444,
"total": 2970,
"trueNegative": 2598,
"truePositive": 232,
"wilsonPrecisionLower95": 0.916
},
"threshold": 0.91
},
"invitation": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8302,
"falseNegative": 24,
"falsePositive": 3,
"precision": 0.9565,
"predictedPositive": 69,
"recall": 0.7333,
"total": 1485,
"trueNegative": 1392,
"truePositive": 66,
"wilsonPrecisionLower95": 0.8798
},
"threshold": 0.935
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.6765,
"falseNegative": 44,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 46,
"recall": 0.5111,
"total": 1485,
"trueNegative": 1395,
"truePositive": 46,
"wilsonPrecisionLower95": 0.9229
},
"threshold": 0.505
}
},
"metrics": {
"f1": 0.5476,
"falseNegative": 111,
"falsePositive": 3,
"precision": 0.9583,
"predictedPositive": 72,
"recall": 0.3833,
"total": 2970,
"trueNegative": 2787,
"truePositive": 69,
"wilsonPrecisionLower95": 0.8845
},
"threshold": 0.935
},
"question": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8445,
"falseNegative": 76,
"falsePositive": 12,
"precision": 0.9522,
"predictedPositive": 251,
"recall": 0.7587,
"total": 1485,
"trueNegative": 1158,
"truePositive": 239,
"wilsonPrecisionLower95": 0.9183
},
"threshold": 0.725
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8536,
"falseNegative": 76,
"falsePositive": 6,
"precision": 0.9755,
"predictedPositive": 245,
"recall": 0.7587,
"total": 1485,
"trueNegative": 1164,
"truePositive": 239,
"wilsonPrecisionLower95": 0.9476
},
"threshold": 0.5
}
},
"metrics": {
"f1": 0.8729,
"falseNegative": 122,
"falsePositive": 26,
"precision": 0.9513,
"predictedPositive": 534,
"recall": 0.8063,
"total": 2970,
"trueNegative": 2314,
"truePositive": 508,
"wilsonPrecisionLower95": 0.9296
},
"threshold": 0.515
},
"replyableMessage": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.948,
"falseNegative": 52,
"falsePositive": 46,
"precision": 0.951,
"predictedPositive": 939,
"recall": 0.945,
"total": 1485,
"trueNegative": 494,
"truePositive": 893,
"wilsonPrecisionLower95": 0.9353
},
"threshold": 0.7
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.9424,
"falseNegative": 62,
"falsePositive": 46,
"precision": 0.9505,
"predictedPositive": 929,
"recall": 0.9344,
"total": 1485,
"trueNegative": 494,
"truePositive": 883,
"wilsonPrecisionLower95": 0.9346
},
"threshold": 0.695
}
},
"metrics": {
"f1": 0.9454,
"falseNegative": 115,
"falsePositive": 90,
"precision": 0.9517,
"predictedPositive": 1865,
"recall": 0.9392,
"total": 2970,
"trueNegative": 990,
"truePositive": 1775,
"wilsonPrecisionLower95": 0.9411
},
"threshold": 0.7
},
"scheduleNegotiation": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 45,
"recall": 1.0,
"total": 1485,
"trueNegative": 1440,
"truePositive": 45,
"wilsonPrecisionLower95": 0.9213
},
"threshold": 0.66
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 45,
"recall": 1.0,
"total": 1485,
"trueNegative": 1440,
"truePositive": 45,
"wilsonPrecisionLower95": 0.9213
},
"threshold": 0.55
}
},
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 90,
"recall": 1.0,
"total": 2970,
"trueNegative": 2880,
"truePositive": 90,
"wilsonPrecisionLower95": 0.9591
},
"threshold": 0.66
},
"task": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8403,
"falseNegative": 89,
"falsePositive": 14,
"precision": 0.9509,
"predictedPositive": 285,
"recall": 0.7528,
"total": 1485,
"trueNegative": 1111,
"truePositive": 271,
"wilsonPrecisionLower95": 0.9192
},
"threshold": 0.835
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8278,
"falseNegative": 98,
"falsePositive": 11,
"precision": 0.9597,
"predictedPositive": 273,
"recall": 0.7278,
"total": 1485,
"trueNegative": 1114,
"truePositive": 262,
"wilsonPrecisionLower95": 0.9293
},
"threshold": 0.745
}
},
"metrics": {
"f1": 0.8637,
"falseNegative": 153,
"falsePositive": 26,
"precision": 0.9562,
"predictedPositive": 593,
"recall": 0.7875,
"total": 2970,
"trueNegative": 2224,
"truePositive": 567,
"wilsonPrecisionLower95": 0.9365
},
"threshold": 0.755
}
}
}
@@ -0,0 +1,629 @@
{
"automaticPromotionAllowed": false,
"baseCorpus": "ModelTraining/ClipboardSemantics/clipboard_semantic_corpus.jsonl",
"configuration": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"configurationsByIntent": {
"blessing": {
"alpha": 1.2e-06,
"augmentation": "numeric",
"char_max": 8,
"char_min": 3,
"external_weight": 0.93,
"hard_example_weight": 2.7,
"l1_ratio": 0.25,
"max_char_features": 95000,
"max_word_features": 45000,
"min_df": 1,
"pseudo_weight": 0.43499999999999994,
"round": 19,
"word_max": 3
},
"complaint": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"confirmationDecision": {
"alpha": 1e-05,
"augmentation": "none",
"char_max": 6,
"char_min": 2,
"external_weight": 0.6100000000000001,
"hard_example_weight": 1.2,
"l1_ratio": 0.05,
"max_char_features": 55000,
"max_word_features": 25000,
"min_df": 2,
"pseudo_weight": 0.195,
"round": 3,
"word_max": 2
},
"followUpReminder": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"invitation": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"question": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"replyableMessage": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"scheduleNegotiation": {
"alpha": 6e-06,
"augmentation": "punctuation",
"char_max": 6,
"char_min": 3,
"external_weight": 0.6900000000000001,
"hard_example_weight": 1.5,
"l1_ratio": 0.1,
"max_char_features": 65000,
"max_word_features": 32000,
"min_df": 2,
"pseudo_weight": 0.255,
"round": 7,
"word_max": 3
},
"task": {
"alpha": 3.5e-06,
"augmentation": "prefix",
"char_max": 6,
"char_min": 2,
"external_weight": 0.8500000000000001,
"hard_example_weight": 2.2,
"l1_ratio": 0.2,
"max_char_features": 85000,
"max_word_features": 42000,
"min_df": 1,
"pseudo_weight": 0.375,
"round": 15,
"word_max": 3
}
},
"openCorpus": "ModelTraining/ClipboardSemantics/open-training-corpus.jsonl",
"reason": "Surrogate feature weights are not deployable NLModel assets and cannot bypass the existing macOS acceptance policy.",
"requiredCommands": [
"python3 Scripts/clipboard_semantics/generate_open_training_corpus.py",
"xcrun swift Scripts/clipboard_semantics/train_models.swift --algorithms maxEnt --corpus ModelTraining/ClipboardSemantics/combined-training-corpus.jsonl",
"xcrun swift Scripts/clipboard_semantics/evaluate_random_holdout.swift --corpus ModelTraining/ClipboardSemantics/random-holdout-corpus.jsonl"
],
"schemaVersion": 1,
"seed": 20260827,
"selectedRound": 1,
"selectedRoundsByIntent": {
"blessing": 19,
"complaint": 1,
"confirmationDecision": 3,
"followUpReminder": 1,
"invitation": 1,
"question": 1,
"replyableMessage": 1,
"scheduleNegotiation": 7,
"task": 15
},
"thresholds": {
"blessing": {
"abstained": true,
"byLanguage": {
"en": {
"abstained": true,
"metrics": {
"f1": 0.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 0.0,
"predictedPositive": 0,
"recall": 0.0,
"total": 0,
"trueNegative": 0,
"truePositive": 0,
"wilsonPrecisionLower95": 0
},
"threshold": 1.0
},
"zh-Hans": {
"abstained": true,
"metrics": {
"f1": 0.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 0.0,
"predictedPositive": 0,
"recall": 0.0,
"total": 0,
"trueNegative": 0,
"truePositive": 0,
"wilsonPrecisionLower95": 0
},
"threshold": 1.0
}
},
"metrics": {
"f1": 0.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 0.0,
"predictedPositive": 0,
"recall": 0.0,
"total": 0,
"trueNegative": 0,
"truePositive": 0,
"wilsonPrecisionLower95": 0
},
"threshold": 1.0
},
"complaint": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 180,
"recall": 1.0,
"total": 1485,
"trueNegative": 1305,
"truePositive": 180,
"wilsonPrecisionLower95": 0.9791
},
"threshold": 0.5
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.9526,
"falseNegative": 9,
"falsePositive": 8,
"precision": 0.9553,
"predictedPositive": 179,
"recall": 0.95,
"total": 1485,
"trueNegative": 1297,
"truePositive": 171,
"wilsonPrecisionLower95": 0.9143
},
"threshold": 0.54
}
},
"metrics": {
"f1": 0.9764,
"falseNegative": 9,
"falsePositive": 8,
"precision": 0.9777,
"predictedPositive": 359,
"recall": 0.975,
"total": 2970,
"trueNegative": 2602,
"truePositive": 351,
"wilsonPrecisionLower95": 0.9567
},
"threshold": 0.54
},
"confirmationDecision": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.9727,
"falseNegative": 1,
"falsePositive": 4,
"precision": 0.957,
"predictedPositive": 93,
"recall": 0.9889,
"total": 1485,
"trueNegative": 1391,
"truePositive": 89,
"wilsonPrecisionLower95": 0.8946
},
"threshold": 0.585
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.7815,
"falseNegative": 31,
"falsePositive": 2,
"precision": 0.9672,
"predictedPositive": 61,
"recall": 0.6556,
"total": 1485,
"trueNegative": 1393,
"truePositive": 59,
"wilsonPrecisionLower95": 0.8881
},
"threshold": 0.5
}
},
"metrics": {
"f1": 0.8701,
"falseNegative": 36,
"falsePositive": 7,
"precision": 0.9536,
"predictedPositive": 151,
"recall": 0.8,
"total": 2970,
"trueNegative": 2783,
"truePositive": 144,
"wilsonPrecisionLower95": 0.9074
},
"threshold": 0.555
},
"followUpReminder": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8554,
"falseNegative": 41,
"falsePositive": 6,
"precision": 0.9586,
"predictedPositive": 145,
"recall": 0.7722,
"total": 1485,
"trueNegative": 1299,
"truePositive": 139,
"wilsonPrecisionLower95": 0.9127
},
"threshold": 0.83
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8354,
"falseNegative": 48,
"falsePositive": 4,
"precision": 0.9706,
"predictedPositive": 136,
"recall": 0.7333,
"total": 1485,
"trueNegative": 1301,
"truePositive": 132,
"wilsonPrecisionLower95": 0.9268
},
"threshold": 0.505
}
},
"metrics": {
"f1": 0.7994,
"falseNegative": 113,
"falsePositive": 11,
"precision": 0.9574,
"predictedPositive": 258,
"recall": 0.6861,
"total": 2970,
"trueNegative": 2599,
"truePositive": 247,
"wilsonPrecisionLower95": 0.9253
},
"threshold": 0.815
},
"invitation": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8302,
"falseNegative": 24,
"falsePositive": 3,
"precision": 0.9565,
"predictedPositive": 69,
"recall": 0.7333,
"total": 1485,
"trueNegative": 1392,
"truePositive": 66,
"wilsonPrecisionLower95": 0.8798
},
"threshold": 0.935
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.6765,
"falseNegative": 44,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 46,
"recall": 0.5111,
"total": 1485,
"trueNegative": 1395,
"truePositive": 46,
"wilsonPrecisionLower95": 0.9229
},
"threshold": 0.505
}
},
"metrics": {
"f1": 0.5476,
"falseNegative": 111,
"falsePositive": 3,
"precision": 0.9583,
"predictedPositive": 72,
"recall": 0.3833,
"total": 2970,
"trueNegative": 2787,
"truePositive": 69,
"wilsonPrecisionLower95": 0.8845
},
"threshold": 0.935
},
"question": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8445,
"falseNegative": 76,
"falsePositive": 12,
"precision": 0.9522,
"predictedPositive": 251,
"recall": 0.7587,
"total": 1485,
"trueNegative": 1158,
"truePositive": 239,
"wilsonPrecisionLower95": 0.9183
},
"threshold": 0.725
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8536,
"falseNegative": 76,
"falsePositive": 6,
"precision": 0.9755,
"predictedPositive": 245,
"recall": 0.7587,
"total": 1485,
"trueNegative": 1164,
"truePositive": 239,
"wilsonPrecisionLower95": 0.9476
},
"threshold": 0.5
}
},
"metrics": {
"f1": 0.8729,
"falseNegative": 122,
"falsePositive": 26,
"precision": 0.9513,
"predictedPositive": 534,
"recall": 0.8063,
"total": 2970,
"trueNegative": 2314,
"truePositive": 508,
"wilsonPrecisionLower95": 0.9296
},
"threshold": 0.515
},
"replyableMessage": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.948,
"falseNegative": 52,
"falsePositive": 46,
"precision": 0.951,
"predictedPositive": 939,
"recall": 0.945,
"total": 1485,
"trueNegative": 494,
"truePositive": 893,
"wilsonPrecisionLower95": 0.9353
},
"threshold": 0.7
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.9424,
"falseNegative": 62,
"falsePositive": 46,
"precision": 0.9505,
"predictedPositive": 929,
"recall": 0.9344,
"total": 1485,
"trueNegative": 494,
"truePositive": 883,
"wilsonPrecisionLower95": 0.9346
},
"threshold": 0.695
}
},
"metrics": {
"f1": 0.9454,
"falseNegative": 115,
"falsePositive": 90,
"precision": 0.9517,
"predictedPositive": 1865,
"recall": 0.9392,
"total": 2970,
"trueNegative": 990,
"truePositive": 1775,
"wilsonPrecisionLower95": 0.9411
},
"threshold": 0.7
},
"scheduleNegotiation": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 45,
"recall": 1.0,
"total": 1485,
"trueNegative": 1440,
"truePositive": 45,
"wilsonPrecisionLower95": 0.9213
},
"threshold": 0.66
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 45,
"recall": 1.0,
"total": 1485,
"trueNegative": 1440,
"truePositive": 45,
"wilsonPrecisionLower95": 0.9213
},
"threshold": 0.55
}
},
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 90,
"recall": 1.0,
"total": 2970,
"trueNegative": 2880,
"truePositive": 90,
"wilsonPrecisionLower95": 0.9591
},
"threshold": 0.66
},
"task": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8593,
"falseNegative": 79,
"falsePositive": 13,
"precision": 0.9558,
"predictedPositive": 294,
"recall": 0.7806,
"total": 1485,
"trueNegative": 1112,
"truePositive": 281,
"wilsonPrecisionLower95": 0.9258
},
"threshold": 0.98
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8166,
"falseNegative": 104,
"falsePositive": 11,
"precision": 0.9588,
"predictedPositive": 267,
"recall": 0.7111,
"total": 1485,
"trueNegative": 1114,
"truePositive": 256,
"wilsonPrecisionLower95": 0.9277
},
"threshold": 0.96
}
},
"metrics": {
"f1": 0.8434,
"falseNegative": 176,
"falsePositive": 26,
"precision": 0.9544,
"predictedPositive": 570,
"recall": 0.7556,
"total": 2970,
"trueNegative": 2224,
"truePositive": 544,
"wilsonPrecisionLower95": 0.934
},
"threshold": 0.965
}
}
}
File diff suppressed because it is too large Load Diff
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,481 @@
{
"automaticPromotionAllowed": false,
"baseCorpus": "ModelTraining/ClipboardSemantics/clipboard_semantic_corpus.jsonl",
"configuration": {
"alpha": 2e-05,
"augmentation": "none",
"char_max": 5,
"char_min": 2,
"external_weight": 0.5700000000000001,
"hard_example_weight": 1.0,
"l1_ratio": 0.0,
"max_char_features": 45000,
"max_word_features": 22000,
"min_df": 2,
"pseudo_weight": 0.16499999999999998,
"round": 1,
"word_max": 2
},
"openCorpus": "ModelTraining/ClipboardSemantics/open-training-corpus.jsonl",
"reason": "Surrogate feature weights are not deployable NLModel assets and cannot bypass the existing macOS acceptance policy.",
"requiredCommands": [
"python3 Scripts/clipboard_semantics/generate_open_training_corpus.py",
"xcrun swift Scripts/clipboard_semantics/train_models.swift --algorithms maxEnt --corpus ModelTraining/ClipboardSemantics/combined-training-corpus.jsonl",
"xcrun swift Scripts/clipboard_semantics/evaluate_random_holdout.swift --corpus ModelTraining/ClipboardSemantics/random-holdout-corpus.jsonl"
],
"schemaVersion": 1,
"seed": 20260827,
"selectedRound": 1,
"thresholds": {
"blessing": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.9318,
"falseNegative": 4,
"falsePositive": 2,
"precision": 0.9535,
"predictedPositive": 43,
"recall": 0.9111,
"total": 1485,
"trueNegative": 1438,
"truePositive": 41,
"wilsonPrecisionLower95": 0.8454
},
"threshold": 0.915
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.7568,
"falseNegative": 17,
"falsePositive": 1,
"precision": 0.9655,
"predictedPositive": 29,
"recall": 0.6222,
"total": 1485,
"trueNegative": 1439,
"truePositive": 28,
"wilsonPrecisionLower95": 0.8282
},
"threshold": 0.9
}
},
"metrics": {
"f1": 0.8375,
"falseNegative": 23,
"falsePositive": 3,
"precision": 0.9571,
"predictedPositive": 70,
"recall": 0.7444,
"total": 2970,
"trueNegative": 2877,
"truePositive": 67,
"wilsonPrecisionLower95": 0.8814
},
"threshold": 0.915
},
"complaint": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 180,
"recall": 1.0,
"total": 1485,
"trueNegative": 1305,
"truePositive": 180,
"wilsonPrecisionLower95": 0.9791
},
"threshold": 0.5
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.9526,
"falseNegative": 9,
"falsePositive": 8,
"precision": 0.9553,
"predictedPositive": 179,
"recall": 0.95,
"total": 1485,
"trueNegative": 1297,
"truePositive": 171,
"wilsonPrecisionLower95": 0.9143
},
"threshold": 0.54
}
},
"metrics": {
"f1": 0.9764,
"falseNegative": 9,
"falsePositive": 8,
"precision": 0.9777,
"predictedPositive": 359,
"recall": 0.975,
"total": 2970,
"trueNegative": 2602,
"truePositive": 351,
"wilsonPrecisionLower95": 0.9567
},
"threshold": 0.54
},
"confirmationDecision": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8848,
"falseNegative": 17,
"falsePositive": 2,
"precision": 0.9733,
"predictedPositive": 75,
"recall": 0.8111,
"total": 1485,
"trueNegative": 1393,
"truePositive": 73,
"wilsonPrecisionLower95": 0.9079
},
"threshold": 0.5
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.7947,
"falseNegative": 30,
"falsePositive": 1,
"precision": 0.9836,
"predictedPositive": 61,
"recall": 0.6667,
"total": 1485,
"trueNegative": 1394,
"truePositive": 60,
"wilsonPrecisionLower95": 0.9128
},
"threshold": 0.5
}
},
"metrics": {
"f1": 0.8418,
"falseNegative": 47,
"falsePositive": 3,
"precision": 0.9779,
"predictedPositive": 136,
"recall": 0.7389,
"total": 2970,
"trueNegative": 2787,
"truePositive": 133,
"wilsonPrecisionLower95": 0.9372
},
"threshold": 0.5
},
"followUpReminder": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8554,
"falseNegative": 41,
"falsePositive": 6,
"precision": 0.9586,
"predictedPositive": 145,
"recall": 0.7722,
"total": 1485,
"trueNegative": 1299,
"truePositive": 139,
"wilsonPrecisionLower95": 0.9127
},
"threshold": 0.83
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8354,
"falseNegative": 48,
"falsePositive": 4,
"precision": 0.9706,
"predictedPositive": 136,
"recall": 0.7333,
"total": 1485,
"trueNegative": 1301,
"truePositive": 132,
"wilsonPrecisionLower95": 0.9268
},
"threshold": 0.505
}
},
"metrics": {
"f1": 0.7994,
"falseNegative": 113,
"falsePositive": 11,
"precision": 0.9574,
"predictedPositive": 258,
"recall": 0.6861,
"total": 2970,
"trueNegative": 2599,
"truePositive": 247,
"wilsonPrecisionLower95": 0.9253
},
"threshold": 0.815
},
"invitation": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8302,
"falseNegative": 24,
"falsePositive": 3,
"precision": 0.9565,
"predictedPositive": 69,
"recall": 0.7333,
"total": 1485,
"trueNegative": 1392,
"truePositive": 66,
"wilsonPrecisionLower95": 0.8798
},
"threshold": 0.935
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.6765,
"falseNegative": 44,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 46,
"recall": 0.5111,
"total": 1485,
"trueNegative": 1395,
"truePositive": 46,
"wilsonPrecisionLower95": 0.9229
},
"threshold": 0.505
}
},
"metrics": {
"f1": 0.5476,
"falseNegative": 111,
"falsePositive": 3,
"precision": 0.9583,
"predictedPositive": 72,
"recall": 0.3833,
"total": 2970,
"trueNegative": 2787,
"truePositive": 69,
"wilsonPrecisionLower95": 0.8845
},
"threshold": 0.935
},
"question": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8445,
"falseNegative": 76,
"falsePositive": 12,
"precision": 0.9522,
"predictedPositive": 251,
"recall": 0.7587,
"total": 1485,
"trueNegative": 1158,
"truePositive": 239,
"wilsonPrecisionLower95": 0.9183
},
"threshold": 0.725
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8536,
"falseNegative": 76,
"falsePositive": 6,
"precision": 0.9755,
"predictedPositive": 245,
"recall": 0.7587,
"total": 1485,
"trueNegative": 1164,
"truePositive": 239,
"wilsonPrecisionLower95": 0.9476
},
"threshold": 0.5
}
},
"metrics": {
"f1": 0.8729,
"falseNegative": 122,
"falsePositive": 26,
"precision": 0.9513,
"predictedPositive": 534,
"recall": 0.8063,
"total": 2970,
"trueNegative": 2314,
"truePositive": 508,
"wilsonPrecisionLower95": 0.9296
},
"threshold": 0.515
},
"replyableMessage": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.948,
"falseNegative": 52,
"falsePositive": 46,
"precision": 0.951,
"predictedPositive": 939,
"recall": 0.945,
"total": 1485,
"trueNegative": 494,
"truePositive": 893,
"wilsonPrecisionLower95": 0.9353
},
"threshold": 0.7
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.9424,
"falseNegative": 62,
"falsePositive": 46,
"precision": 0.9505,
"predictedPositive": 929,
"recall": 0.9344,
"total": 1485,
"trueNegative": 494,
"truePositive": 883,
"wilsonPrecisionLower95": 0.9346
},
"threshold": 0.695
}
},
"metrics": {
"f1": 0.9454,
"falseNegative": 115,
"falsePositive": 90,
"precision": 0.9517,
"predictedPositive": 1865,
"recall": 0.9392,
"total": 2970,
"trueNegative": 990,
"truePositive": 1775,
"wilsonPrecisionLower95": 0.9411
},
"threshold": 0.7
},
"scheduleNegotiation": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 1.0,
"falseNegative": 0,
"falsePositive": 0,
"precision": 1.0,
"predictedPositive": 45,
"recall": 1.0,
"total": 1485,
"trueNegative": 1440,
"truePositive": 45,
"wilsonPrecisionLower95": 0.9213
},
"threshold": 0.5
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.967,
"falseNegative": 1,
"falsePositive": 2,
"precision": 0.9565,
"predictedPositive": 46,
"recall": 0.9778,
"total": 1485,
"trueNegative": 1438,
"truePositive": 44,
"wilsonPrecisionLower95": 0.8547
},
"threshold": 0.93
}
},
"metrics": {
"f1": 0.978,
"falseNegative": 1,
"falsePositive": 3,
"precision": 0.9674,
"predictedPositive": 92,
"recall": 0.9889,
"total": 2970,
"trueNegative": 2877,
"truePositive": 89,
"wilsonPrecisionLower95": 0.9085
},
"threshold": 0.91
},
"task": {
"abstained": false,
"byLanguage": {
"en": {
"abstained": false,
"metrics": {
"f1": 0.8403,
"falseNegative": 89,
"falsePositive": 14,
"precision": 0.9509,
"predictedPositive": 285,
"recall": 0.7528,
"total": 1485,
"trueNegative": 1111,
"truePositive": 271,
"wilsonPrecisionLower95": 0.9192
},
"threshold": 0.835
},
"zh-Hans": {
"abstained": false,
"metrics": {
"f1": 0.8278,
"falseNegative": 98,
"falsePositive": 11,
"precision": 0.9597,
"predictedPositive": 273,
"recall": 0.7278,
"total": 1485,
"trueNegative": 1114,
"truePositive": 262,
"wilsonPrecisionLower95": 0.9293
},
"threshold": 0.745
}
},
"metrics": {
"f1": 0.8637,
"falseNegative": 153,
"falsePositive": 26,
"precision": 0.9562,
"predictedPositive": 593,
"recall": 0.7875,
"total": 2970,
"trueNegative": 2224,
"truePositive": 567,
"wilsonPrecisionLower95": 0.9365
},
"threshold": 0.755
}
}
}
@@ -0,0 +1,129 @@
{
"schemaVersion": 1,
"completedResearchRounds": 60,
"requestedCoreRounds": 20,
"seed": 20260827,
"data": {
"baseTrainingRecords": 11500,
"reachableLicensedOpenRecords": 3845,
"holdoutOverlapCounts": {
"comprehensiveOnline": 0,
"random": 0,
"targetedRelease": 0
},
"unavailableSources": [
"MASSIVE",
"CLINC150",
"CFPB"
]
},
"phases": [
{
"name": "shared-configuration-search",
"rounds": 20,
"selectedRound": 1,
"golden": {
"macroPrecision": 0.9366,
"macroRecall": 0.6801,
"macroF1": 0.7711
},
"randomHoldout": {
"macroPrecision": 0.8569,
"macroRecall": 0.4603,
"macroF1": 0.5709
},
"researchOnlyComprehensive": {
"macroPrecision": 0.4398,
"macroRecall": 0.2161,
"macroF1": 0.2383
},
"targetedRelease": {
"macroPrecision": 0.5489,
"macroRecall": 0.3139,
"macroF1": 0.3899
}
},
{
"name": "per-intent-fine-tuning",
"rounds": 20,
"selectedRoundsByIntent": {
"blessing": 19,
"complaint": 1,
"confirmationDecision": 3,
"followUpReminder": 1,
"invitation": 1,
"question": 1,
"replyableMessage": 1,
"scheduleNegotiation": 7,
"task": 15
},
"golden": {
"macroPrecision": 0.9377,
"macroRecall": 0.6819,
"macroF1": 0.7666
},
"randomHoldout": {
"macroPrecision": 0.8701,
"macroRecall": 0.4705,
"macroF1": 0.5770
},
"researchOnlyComprehensive": {
"macroPrecision": 0.4292,
"macroRecall": 0.2250,
"macroF1": 0.2645
},
"targetedRelease": {
"macroPrecision": 0.5467,
"macroRecall": 0.3375,
"macroF1": 0.4095
}
},
{
"name": "explicit-evidence-fine-tuning",
"rounds": 20,
"selectedRound": 1,
"golden": {
"macroPrecision": 0.9877,
"macroRecall": 0.4354,
"macroF1": 0.5476
},
"randomHoldout": {
"macroPrecision": 0.7731,
"macroRecall": 0.3029,
"macroF1": 0.3905
},
"researchOnlyComprehensive": {
"macroPrecision": 0.6097,
"macroRecall": 0.1647,
"macroF1": 0.2325
},
"targetedRelease": {
"macroPrecision": 0.5489,
"macroRecall": 0.2537,
"macroF1": 0.3209
}
}
],
"currentDeployedReference": {
"randomHoldout": {
"macroPrecision": 0.9832,
"macroRecall": 0.6557,
"macroF1": 0.7669
},
"researchOnlyComprehensive": {
"macroPrecision": 0.2948,
"macroRecall": 0.3320,
"macroF1": 0.2783
}
},
"findings": [
"The first low-complexity round generalized better than later self-training rounds.",
"Three-round high-confidence pseudo-labeling amplified source and template bias instead of improving frozen regressions.",
"Per-intent configuration selection improved the research-only comprehensive F1 from 0.2383 to 0.2645 but remained below the deployed reference F1 of 0.2783.",
"Explicit evidence gates raised golden precision to 0.9877 and research-only precision to 0.6097, but recall collapsed and the release gate still failed.",
"No surrogate candidate is eligible to replace or augment the deployed Apple models."
],
"releaseGatePassed": false,
"deploymentDecision": "retain-current-deployed-models",
"platformLimitation": "Linux cannot train, compile, run, or benchmark the deployable Apple Create ML models. These rounds are surrogate research and produce a macOS replay specification, not app-ready model assets."
}
File diff suppressed because it is too large Load Diff