Speculative decoding
SVIP
Superseded baseline#37 of 151 most-superseded
Superseded — cited as a baseline and beaten by newer methods
2 papers critique it · 1 beat it on benchmarks
What papers say
Verbatim critique sentences, each from a paper that cites SVIP as a baseline.
However, unlike , none of these techniques focus on the data movement cost due to speculation. They require access to output probability distributions and are incompatible with approaches like n-gram speculation.
“a static threshold across all prompts and positions in the drafting sequence, such as the one used in SVIP, is suboptimal”
Beaten on benchmarks
Head-to-head results where a newer method reports beating SVIP. Values are copied from the source paper's tables — verify against the cited paper.
TapOut - Seq UCB1 beats SVIP
1.22 vs 0.68
speedup · [Llama-3 1B/8B]
TapOut: A Bandit-Based Approach to Dynamic Speculative Decoding
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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.
- Nov 3, 2025