On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes
This addresses a key bottleneck in compressing large language models for practical deployment, though it is an incremental improvement over existing distillation methods.
The paper tackles the distribution mismatch problem in knowledge distillation for auto-regressive sequence models by introducing Generalized Knowledge Distillation (GKD), which trains the student on self-generated outputs with teacher feedback, and demonstrates its efficacy on tasks like summarization, translation, and arithmetic reasoning.
Knowledge distillation (KD) is widely used for compressing a teacher model to reduce its inference cost and memory footprint, by training a smaller student model. However, current KD methods for auto-regressive sequence models suffer from distribution mismatch between output sequences seen during training and those generated by the student during inference. To address this issue, we introduce Generalized Knowledge Distillation (GKD). Instead of solely relying on a fixed set of output sequences, GKD trains the student on its self-generated output sequences by leveraging feedback from the teacher on such sequences. Unlike supervised KD approaches, GKD also offers the flexibility to employ alternative loss functions between the student and teacher, which can be useful when the student lacks the expressivity to mimic the teacher's distribution. Furthermore, GKD facilitates the seamless integration of distillation with RL fine-tuning (RLHF). We demonstrate the efficacy of GKD for distilling auto-regressive language models on summarization, translation, and arithmetic reasoning tasks, and task-agnostic distillation for instruction-tuning.