Training Data

In one sentenceThe text, images or other examples a model learned from. It shapes what the model knows and how it behaves.

What it means

Before a model can answer anything, it is trained on a huge collection of examples. For an LLM that means a large slice of the public internet, books, code and licensed data, followed by extra training with human feedback to make it helpful and safe.

What is in the training data determines the model's strengths, its blind spots and its biases. It also sets the Knowledge Cutoff.

How to use it

• Check each tool's settings to see whether your chats are used to train future models, and turn it off for sensitive work if you need to.
• Expect weaker answers on niche topics that were rare on the internet. Give the model your own material instead.

Related terms

Knowledge Cutoff AI Bias Fine-Tuning AI Data Privacy