Out-of-Distribution

Definitions

1. Describing inputs or situations that differ meaningfully from the data or conditions a model encountered during training. Out-of-distribution cases are important because model behavior can degrade unpredictably outside familiar patterns.

   In plain English: Different from the kinds of data or situations the model was trained on.

   Example: "The classifier looked strong on test data but failed badly on out-of-distribution scans from a new document vendor."

Related Terms

• Overfitting
• Underfit
• AI Multi-Agent
• Alignment Tax
• Benchmark Saturation
• Goal Misgeneralization
• Robustness
• Zero-Shot