Instructions to use SkillScreen/SkillScreen-2B-Package with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SkillScreen/SkillScreen-2B-Package with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="SkillScreen/SkillScreen-2B-Package")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForSequenceClassification processor = AutoProcessor.from_pretrained("SkillScreen/SkillScreen-2B-Package") model = AutoModelForSequenceClassification.from_pretrained("SkillScreen/SkillScreen-2B-Package", device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 4,515 Bytes
4fa3363 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 | {
"training_scope": "all_data",
"seed": 42,
"input_mode": "package32k",
"max_length": 32768,
"hypothesis": "This AI agent skill is malicious.",
"dataset": {
"repo_id": "SkillScreen/SkillScreen-Data",
"revision": "d59aabbd7e0e8d95b464a2d335f36e948577b832"
},
"native_score": "entailment softmax",
"thresholds": {
"mode": "package32k",
"seed": 42,
"records": 11831,
"malicious": 6436,
"benign": 5395,
"auprc_average_precision": 0.9990312216598736,
"auroc": 0.9984704751021387,
"folds_sha256": "c11c6f68dd23fddb70d7090512d93f5be54bb216596175d44fcaa7e7df80e202",
"oof_sha256": "f46a75a97b3716c787bd7116fb837ee0124be34055c1dbce75e6f6d1c27933f9",
"folds": [
{
"fold": 1,
"predictions_sha256": "e9e07bbeca1b9550d997de45dad000527ee6e212cf140247b7ab5885691aa6eb",
"adapter_sha256": "1407f6319076816ea490e8e11b139c01ec3ff80947bd8fa962e16a57488eb85f"
},
{
"fold": 2,
"predictions_sha256": "94ba50e421d4385b7c0fd42cf6f1da076df7f81ec4b3e2a2872a21e015afb2b0",
"adapter_sha256": "a01b9e8ab0623a54a7ae845137fb4fa0c877c9ddda6c57ac00d1600e0caafb1c"
},
{
"fold": 3,
"predictions_sha256": "f007afc29a1b5aa7407287501fc52619bb56fd1340197874aae5b5faccb50e14",
"adapter_sha256": "e2261670e1f652da8bc316b3449c8ee03b0f5d993b7b958ddeb7b89a638bf116"
}
],
"created_at": "2026-11-10T07:34:36.103352+00:00",
"default": {
"raw_threshold": 0.0,
"native_threshold": 0.5,
"tp": 6367,
"fp": 18,
"fn": 69,
"tn": 5377,
"recall": 0.9892790553138595,
"fpr": 0.003336422613531047,
"precision": 0.9971808927173061,
"f1": 0.9932142578582014,
"flagged_fraction": 0.5396838813287127
},
"targets": {
"0.95": {
"raw_threshold": 7.327948239240634,
"native_threshold": 0.9993435112125237,
"tp": 6115,
"fp": 1,
"fn": 321,
"tn": 5394,
"recall": 0.9501243008079553,
"fpr": 0.00018535681186283596,
"precision": 0.999836494440811,
"f1": 0.974346717654557,
"flagged_fraction": 0.5169470036345195
},
"0.98": {
"raw_threshold": 5.646373107267836,
"native_threshold": 0.9964821215635228,
"tp": 6308,
"fp": 4,
"fn": 128,
"tn": 5391,
"recall": 0.9801118707271598,
"fpr": 0.0007414272474513438,
"precision": 0.9993662864385298,
"f1": 0.989645434577973,
"flagged_fraction": 0.5335136505789874
},
"0.99": {
"raw_threshold": -0.9062583981072989,
"native_threshold": 0.2877660945090445,
"tp": 6372,
"fp": 19,
"fn": 64,
"tn": 5376,
"recall": 0.9900559353635798,
"fpr": 0.0035217794253938834,
"precision": 0.997027069316226,
"f1": 0.9935292741872612,
"flagged_fraction": 0.5401910235821148
}
},
"balanced": {
"target": 0.99,
"raw_threshold": -0.9062583981072989,
"native_threshold": 0.2877660945090445,
"tp": 6372,
"fp": 19,
"fn": 64,
"tn": 5376,
"recall": 0.9900559353635798,
"fpr": 0.0035217794253938834,
"precision": 0.997027069316226,
"f1": 0.9935292741872612,
"flagged_fraction": 0.5401910235821148
},
"high_recall": {
"raw_threshold": -0.9062583981072989,
"native_threshold": 0.2877660945090445,
"tp": 6372,
"fp": 19,
"fn": 64,
"tn": 5376,
"recall": 0.9900559353635798,
"fpr": 0.0035217794253938834,
"precision": 0.997027069316226,
"f1": 0.9935292741872612,
"flagged_fraction": 0.5401910235821148
},
"selection": "Balanced maximizes recall minus FPR among predeclared 95/98/99% OOF recall gates; ties prefer higher recall. High-recall targets 99% OOF recall.",
"scope": "Pooled three-fold out-of-fold scores. Every score is from a model that excluded that skill group. Thresholds transfer to a separate full-data refit; these are not held-out test results of that refit."
},
"metric_scope": "OOF metrics belong to the fold models, not an independent test of this final refit.",
"repo_id": "SkillScreen/SkillScreen-2B-Package",
"release": "v0.1.0",
"default_policy": "high-recall",
"base_model": {
"repo_id": "AlexWortega/openjev",
"subfolder": "qwen3.5-2b-nli-v5",
"revision": "a20448012c213128955ca0c693e7c943865cab77"
}
}
|