A type of machine learning that relies on generating implicit la­bels from unstructured data rather than relying on explicit, human-created labels. Self-supervised learning tasks are constructed to allow the true labels to be automatically inferred from the training data (enabling the use of large-scale training data) and to require models to capture essential features or relationships within the data to solve them. For example, a common self-supervised learning task is providing a model with partial data with the task to accurately generate the re­mainder.
Sources:
NIST AI 100-2e2025