Preprint
Nov 2026
SpecPilot: Autonomous Algorithm Discovery and Systems Optimization for Speculative Decoding
Speculative decoding accelerates memory-bound LLM decoding with a lightweight drafter, and parallel drafters extend this speedup by drafting a whole block in one forward pass, but it remains unclear whether their released systems attain the full speedup that parallel drafting permits.
Kun-Jun Li, Wen-Hao Chai, Yihao Liang et al.
· 0 citations