# What is Contrastive Learning? Definition & Technical Mechanics

> **Source:** [SPIDITS AI AI Glossary](https://spidits.com/ai-glossary)  
> **Category:** Model Training  
> **Canonical Citation:** [https://spidits.com/ai-glossary/contrastive-learning](https://spidits.com/ai-glossary/contrastive-learning)

## Definition
Contrastive Learning is a self-supervised training technique where a model learns to group similar inputs (positive pairs) close together in embedding space while pushing dissimilar inputs (negative pairs) far apart.

## Architectural Mechanics & Deep Dive
Contrastive learning is a self-supervised learning paradigm where a model learns data representations by comparing similar and dissimilar samples. The training objective forces representations of positive pairs (e.g., two different views of the same image) to be close together in vector space, while pushing representations of negative pairs (different images) far apart, enabling the model to learn semantic features without labels.

## Industry Applications & Real-World Use Cases
Contrastive Image-Text Pre-training (CLIP), sentence embeddings, and visual representations

## Frequently Asked Questions

### How is contrastive learning used in multimodal AI?
It aligns images and their text captions by training the model to match the correct image-text pairs in a shared embedding space.

### What is a positive vs. negative pair?
A positive pair consists of different representations of the same concept (e.g. cropped versions of the same photo). A negative pair contains unrelated concepts.

## Related Terminology
- [embedding](https://spidits.com/ai-glossary/embedding)
- [multimodal-ai](https://spidits.com/ai-glossary/multimodal-ai)
- [unsupervised-learning](https://spidits.com/ai-glossary/unsupervised-learning)

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