Target Regeneration Shrinks the Gains of Draft-Training Objectives in Speculative Decoding
Speculative decoding accelerates large language models by verifying drafted tokens in parallel. New draft-training objectives are usually credited with the acceptance they add over cross-entropy on externally written responses, yet many pipelines let the target rewrite them.
By Jungseob Lee, Hogun Seong, Euijun Choi, Sugyeong EoarXiv - cs.CL (NLP)
Read the full story
Originally published by arXiv - cs.CL (NLP) on Oct 22, 2026.
Comments
Sign in to join the discussion.