# What is Out-of-Distribution? Definition & Technical Mechanics

> **Source:** [SPIDITS AI AI Glossary](https://spidits.com/ai-glossary)  
> **Category:** Model Limitations  
> **Canonical Citation:** [https://spidits.com/ai-glossary/out-of-distribution](https://spidits.com/ai-glossary/out-of-distribution)

## Definition
Out-of-Distribution (OOD) data refers to inputs that originate from a different probability distribution than the dataset used to train the machine learning model, often causing models to make confident mistakes.

## Architectural Mechanics & Deep Dive
Out-of-distribution (OOD) data refers to inputs processed during inference whose statistical distribution differs significantly from the training dataset. Machine learning models struggle to make accurate predictions on OOD data because their generalization assumptions fail, driving research into anomaly detection and model robustness to prevent silent failures in production.

## Industry Applications & Real-World Use Cases
Robustness validation runs, anomaly detection systems, and model safety checks.

## Frequently Asked Questions

### Why do models fail on out-of-distribution inputs?
Because models make predictions based on statistical correlation patterns learned during training, which do not hold true for fundamentally different data.

### How do you detect out-of-distribution inputs?
By monitoring embedding distances or using classification confidence thresholds to route anomalous inputs to safety loops.

## Related Terminology
- [overfitting](https://spidits.com/ai-glossary/overfitting)
- [adversarial-attack](https://spidits.com/ai-glossary/adversarial-attack)
- [hallucination](https://spidits.com/ai-glossary/hallucination)

---
*From the SPIDITS AI Knowledge Base (336+ terms indexed). Explore full technical definitions: [https://spidits.com/ai-glossary](https://spidits.com/ai-glossary)*