Which learning approach works by adjusting behavior to labels and features in a test set generated automatically?

Which learning approach works by adjusting behavior to labels and features in a test set generated automatically?

• A. Supervised learning
• B. Self-supervised learning
• C. Mask learning
• D. Generative AI

Answer: (B)

Self-supervised learning uses supervisory signals created from the data itself rather than relying on human-labeled examples. It designs a task where part of the input is transformed, masked, or shuffled, and the model must predict the original or the missing pieces. The labels come from the data generation process, so the model learns by adjusting its behavior to these automatically produced targets. This matches the idea of a test set generated automatically, since the targets are created without external labeling. Supervised learning needs human-provided labels, so it doesn’t fit the automatic-label scenario. Mask learning is a specific self-supervised technique, not the broad approach itself, and generative AI focuses more on producing new data than on learning from automatically generated labels.

Self-supervised learning uses supervisory signals created from the data itself rather than relying on human-labeled examples. It designs a task where part of the input is transformed, masked, or shuffled, and the model must predict the original or the missing pieces. The labels come from the data generation process, so the model learns by adjusting its behavior to these automatically produced targets. This matches the idea of a test set generated automatically, since the targets are created without external labeling. Supervised learning needs human-provided labels, so it doesn’t fit the automatic-label scenario. Mask learning is a specific self-supervised technique, not the broad approach itself, and generative AI focuses more on producing new data than on learning from automatically generated labels.