Trajectory as the Teacher: Few-Step Discrete Flow Matching via Energy-Navigated Distillation

Researchers at Apple have introduced a method called Trajectory as the Teacher to accelerate discrete flow matching for text generation. By using the multi-step paths of a complex model to train a smaller student model, this technique reduces the number of forward passes required to produce coherent language. This approach aims to maintain output quality while significantly lowering the computational resources needed for text inference.