PELM: Power Efficient On-Device LLM Inference with Speculative Decoding and Dynamic Voltage Frequency Scaling.

Weisi Yang, Stephen Xia

Browse the full SENSYS paper archive.