# What is Model Collapse? Definition & Technical Mechanics

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

## Definition
Model Collapse is a degenerative process affecting generative AI models trained recursively on synthetic data generated by previous generations of AI models. Over iterations, the model loses diversity, starts repeating patterns, and eventually outputs garbage.

## Architectural Mechanics & Deep Dive
Model collapse is a degenerative process where an AI model trained on synthetic data (data generated by other AI models) begins to lose its capability, forget rare data distributions, and generate repetitive, low-quality outputs. This occurs as errors accumulate over generations, highlighting the critical value of human-curated datasets.

## Industry Applications & Real-World Use Cases
Dataset curation audits, synthetic data quality checks, and web scraping filters.

## Frequently Asked Questions

### Why does model collapse happen?
Statistical approximation errors build up in each training generation. Rare events and tail data distributions are filtered out, leading to statistical decay.

### How can model collapse be avoided?
By ensuring training datasets contain a strong baseline of human-created data and filtering out low-quality synthetic inputs.

## Related Terminology
- [synthetic-data](https://spidits.com/ai-glossary/synthetic-data)
- [overfitting](https://spidits.com/ai-glossary/overfitting)
- [generative-ai](https://spidits.com/ai-glossary/generative-ai)

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