Beyond Autoregression: Fast LLMs via Self-Distillation Through Time

Open in new window