Curriculum learning trains a model on easier examples first and harder ones later, the way a school course builds up.
In plain words
Most training shows examples in random order. Curriculum learning sorts them by difficulty and gradually adds the hard ones. Two choices shape it: how to judge an example's difficulty, and how quickly to introduce harder data. It can help a model learn faster or end up more general.
A quick example
To teach a model maths word problems, a team first trains on one-step sums, then on two-step problems, and only later on long multi-step questions. The model builds on what it has already learned instead of meeting the hardest cases on day one.
Where you'll meet it
- Image recognition and language models.
- Helping a model learn faster or generalise better.
Related
Sources
- PaperA Survey on Curriculum Learning, Wang, Chen and Zhu (arXiv) · read 29 Sept 2026
- PaperCurriculum Learning: A Survey, Soviany et al. (arXiv) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.