# Training-Time-Only Model Logic

> Building logic that exists only during training, with no counterpart to serve predictions at inference.

Record: `lexicon:training-time-only-model-logic`
Kind: anti-pattern
Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md)
Canonical: https://banes-lab.com/ontology/lexicon#lexicon-training-time-only-model-logic

Listed in [Lexicon terms](https://banes-lab.com/api/records/lexicon.md), after [Opaque Ungoverned Model Use](https://banes-lab.com/records/lexicon/opaque-ungoverned-model-use.md) and before [Unapproved Model Deployment](https://banes-lab.com/records/lexicon/unapproved-model-deployment.md).

## Category

- [Model Architecture](https://banes-lab.com/records/lexicon-category/model-architecture.md)

## Negated by

- [Model Inference](https://banes-lab.com/records/architecture/model-inference.md)

## Linked from

- [Model Architecture](https://banes-lab.com/ontology/principles/architecture-category-model-architecture.md)