Definition
Curriculum learning is a training strategy that feeds a model examples in a deliberately chosen order, typically from easy to hard, rather than in random order. The idea, drawn from how humans are taught, is that mastering simple cases first builds a foundation that makes the harder cases learnable. Formalized for machine learning by Bengio and colleagues in 2009, it can speed up convergence and, on some tasks, lift the final quality the model reaches.
How a curriculum is built
The core ingredients are a difficulty measure that scores how hard each example is, and a pacing function that decides how quickly harder material is introduced. Difficulty might be sentence length, problem complexity, or how confidently a reference model handles the example. Training begins on the easy subset and gradually admits harder examples until the full distribution is in play. Variants include anti-curriculum (hard first) and self-paced learning, where the model itself helps judge which examples it is ready for.
Why it matters
For builders training or fine-tuning models on limited hardware, a good curriculum can mean reaching a target quality in fewer steps, which directly saves compute and time. It also tends to stabilize early training, when a flood of hard examples can send a model into poor regions of the loss landscape. Curriculum learning is a scheduling choice rather than an architecture change, so it composes with the rest of the training stack.
Ordering interacts with how long you train — see training epoch — and with the broader hunt for good hyperparameters.
In Simple Terms
Curriculum learning is a training strategy that feeds a model examples in a deliberately chosen order, typically from easy to hard, rather than in random…