Training Efficiency

AdaSTaR: Adaptive Data Sampling for Training Self-Taught Reasoners