Technical
Model artefacts
Learned model components such as weights or pre-trained models.
Definition
Refers to components produced during model development, such as trained weights, pre-trained models, and other derived outputs of the training process. The guidance treats these artefacts as distinct from raw training data and as items that can engage privacy obligations (for example via leakage or provenance concerns).
Related Terms
Trained model
The model artifact produced by training on a dataset....
Models
Trained algorithms used to generate inferences or predictions....
Training dataset
Dataset used to train a machine learning model....
Model
A trained algorithmic artifact that performs tasks using learned parameters....
Critical Technical Artifacts (CTAs)
Model weights, training data, or code that enable capability replication....