1 paper · 1 filter
Rui Wei, Rui Du, Hanfei Yu +4
In Large Language Model (LLM) inference, early-exit refers to stopping computation at an intermediate layer once the prediction is sufficiently confident, thereby reducing latency…