Skip to content
AImpact
IT EN
Data Beginner Also known as: Dati sintetici

Synthetic Data

Training data generated by another AI model instead of collected from humans.

ShareLinkedInX

In practice

It is now a pillar of modern training: big models produce examples to train smaller ones (distillation) or to cover rare cases. It must be filtered carefully, because generator errors compound in the final model. Nvidia, Meta, and Anthropic use it heavily.

Related terms

Pretraining SFT Distillation Fine-tuning

Seen in the wild

5 entries mentioning it
  1. NVIDIA Isaac GR00T N1: the open foundation model for humanoid robots unveiled at GTC
    High
  2. Nemotron-4 340B: NVIDIA's model family for generating synthetic training data
    Medium
  3. RoboGen: an agent that generates its own tasks, scenes and supervision to train robots in simulation
    Medium
  4. Phi-1.5: 1.3 billion parameters trained on textbook-style synthetic data
    High
  5. Phi-1: 1.3 billion parameters and 50.6% on HumanEval
    High
← All terms