The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.
YAML Metadata Warning:The task_ids "text-classification" is not in the official list: acceptability-classification, entity-linking-classification, fact-checking, intent-classification, language-identification, multi-class-classification, multi-label-classification, multi-input-text-classification, natural-language-inference, semantic-similarity-classification, sentiment-classification, topic-classification, semantic-similarity-scoring, sentiment-scoring, sentiment-analysis, hate-speech-detection, text-scoring, named-entity-recognition, part-of-speech, parsing, lemmatization, word-sense-disambiguation, coreference-resolution, extractive-qa, open-domain-qa, closed-domain-qa, news-articles-summarization, news-articles-headline-generation, dialogue-modeling, dialogue-generation, conversational, language-modeling, text-simplification, explanation-generation, abstractive-qa, open-domain-abstractive-qa, closed-domain-qa, open-book-qa, closed-book-qa, text2text-generation, slot-filling, masked-language-modeling, keyword-spotting, speaker-identification, audio-intent-classification, audio-emotion-recognition, audio-language-identification, multi-label-image-classification, multi-class-image-classification, face-detection, vehicle-detection, instance-segmentation, semantic-segmentation, panoptic-segmentation, image-captioning, image-inpainting, image-colorization, super-resolution, grasping, task-planning, tabular-multi-class-classification, tabular-multi-label-classification, tabular-single-column-regression, rdf-to-text, multiple-choice-qa, multiple-choice-coreference-resolution, document-retrieval, utterance-retrieval, entity-linking-retrieval, fact-checking-retrieval, univariate-time-series-forecasting, multivariate-time-series-forecasting, visual-question-answering, document-question-answering, pose-estimation
Permission Command Dataset
Synthetic and augmented command-line examples for training a local permission classifier that helps gate command execution in an application runtime.
Status, 30 August 2026
The two command models trained on this corpus were degenerate and have been withdrawn. Each emitted the same score for every input, and the metrics shipped beside them report ROC AUC 0.5000, which is what a constant scores rather than a measurement of skill.
They were command_hazard_model.onnx and obfuscation_model.onnx in
cowWhySo/permission-gate-onnx. Both remain at the tag v0-degenerate-models
in that repository, and its card carries the full measurement.
The tell is in the published numbers. Every average_precision equals its class
base rate to the last digit, and that happens only when every score is tied:
| Model and split | Reported AP | Base rate |
|---|---|---|
| hazard, validation | 0.5142857142857142 | 18/35 |
| hazard, test | 0.325 | 13/40 |
| obfuscation, validation | 0.42857142857142855 | 15/35 |
| obfuscation, test | 0.525 | 21/40 |
The cause is feature scaling in the training notebooks. Raw len(s) and raw
character counts were appended to two L2-normalized hashing blocks, so one column
carried magnitude in the thousands beside two blocks whose entire norm was 1.0.
Training then used alpha=1e-6 under learning_rate="optimal", which derives
the step size from alpha. That asked for a very large step and no
regularization at once, the weights diverged, and the sigmoid saturated.
Bounding those features and raising alpha to 1e-4 fixes it. The same
pipeline then reaches test ROC AUC 0.9972 for the hazard model and 0.9774 for the
obfuscation model.
The two notebooks confirm this between them, without meaning to. Both train with
alpha=1e-6. The ONNX notebook's build_command_sparse_features appends the
lexical block and produces the same ROC AUC of 0.5000, so its exported
command_hazard_model.onnx and obfuscation_model.onnx are dead too. Its
build_pi_sparse_features omits that block and returns
hstack([x_char, x_word]) alone, and that model scores 0.9811.
One column is the only difference between those two outcomes. That is a controlled comparison sitting in the published artifacts already.
The fixed numbers still need a caveat
Held out by generator, the hazard model falls from 0.9972 to 0.7010. 181 of the 246 training rows come from a single model, so much of the in-distribution score is the classifier recognizing that model's phrasing. Report the held-out number alongside the pooled one.
See the permission-command-corpus card for the split statistics and for what
this corpus does not answer, and the prompt-injection-watch-dataset card for a
larger source confound found in the sibling corpus.
Labels
actual_label: ground-truth good versus badsurface_label: whether the command appears benign or suspicious at a glanceobfuscation_label: plain versus obfuscatedobfuscation_type: specific obfuscation familyrisk_class: broad behavioral category
Splits
| Split | Rows | Positive rate |
|---|---|---|
| train | 246 | 0.504 |
| validation | 35 | 0.514 |
| test | 40 | 0.325 |
Caveats
- Contains synthetic examples
- Contains LLM-generated examples, 181 of 246 training rows from one generator
- The test split is 40 rows, which is too few to separate models that are close
- Must be reviewed and expanded with real-world false positives and false negatives before production use
- Downloads last month
- 18