Detailed explanation
Model quality starts with the dataset: size, diversity, cleanliness, low bias and good labeling. We usually split it into training, validation and test sets to measure generalization.
Examples
ImageNet for vision
Common Crawl for web text
Internal ticket data to fine-tune a support chatbot
Frequently asked questions
How big should a dataset be?
A few hundred examples can be enough for fine-tuning; foundation models train on terabytes.
Related terms
Embedding
An embedding is a numerical vector representation of text, image, or audio data that captures its semantic meaning, enabling similarity search and machine learning on unstructured content.
Fine-tuning
Fine-tuning is the process of further training a pre-trained model on a smaller, task-specific dataset to specialize it for a particular domain, style, or task.
Machine Learning
A branch of AI where models learn patterns from data instead of following explicit rules.
Supervised Learning
Learning from labeled examples (input → expected output).
Last updated: 7/15/2026