Scholar Feed

Speculative decoding

AdaSpec

AdaSPEC: Selective Knowledge Distillation for Efficient Speculative Decoders

Superseded baseline#84 of 151 most-superseded · first seen Oct 22, 2025

Cited as a baseline — critiqued by newer work, not yet beaten on a benchmark here

1 papers critique it · 0 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites AdaSpec as a baseline.

AdaSpec designs dynamic length tuning, but still follows a round-based speculate-then-verify workflow and therefore cannot hide draft latency behind target execution.
— FASER: Fine-Grained Phase Management for Speculative Decoding in Dynamic LLM Serving

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.