[short] Think Big, Generate Quick: LLM-to-SLM for Fast Autoregressive Decoding

Arxiv Papers - [short] Think Big, Generate Quick: LLM-to-SLM for Fast Autoregressive Decoding