Level Up: Defining and Exploiting Transitional Problems for Curriculum Learning
This addresses the challenge of designing effective curricula for machine learning, offering a more efficient and interpretable approach compared to static or dynamic methods, though it is incremental in improving existing curriculum learning techniques.
The paper tackled the problem of inefficient curriculum learning by introducing a method to measure problem difficulty directly relative to model ability, identifying transitional problems that become easier as the model improves. In chess and mathematics, training on a curriculum based on these problems outperformed other strategies, efficiently advancing models to higher competence tiers.
Curriculum learning--ordering training examples in a sequence to aid machine learning--takes inspiration from human learning, but has not gained widespread acceptance. Static strategies for scoring item difficulty rely on indirect proxy scores of varying quality and produce curricula that are not specific to the learner at hand. Dynamic approaches base difficulty estimates on gradient information, requiring considerable extra computation during training. We introduce a novel method for measuring the difficulty of individual problem instances directly relative to the ability of a given model, and identify transitional problems that are consistently easier as model ability increases. Applying this method to chess and mathematics, we find that training on a curriculum that "levels up" from easier to harder transitional problems most efficiently improves a model to the next tier of competence. These problems induce a natural progression from easier to harder items, which outperforms other training strategies. By measuring difficulty directly relative to model competence, our method yields interpretable problems, learner-specific curricula, and a principled basis for step-by-step improvement.