HASS (Speculative decoding): superseded — cited as a baseline and beaten by newer methods. 2 paper(s) critique it, 6 beat it on benchmarks — #13 of 151 most-superseded. Sub-problem: cluster led by EAGLE-2. Newer alternatives in the same sub-problem include DREAM-S, PPOW, SpecForge, OnlineSpec, MoE-Spec.

Superseded baseline#13 of 151 most-superseded

HASS

Speculative decoding

superseded — cited as a baseline and beaten by newer methods

2 papers critique it · 6 beat it on benchmarks

What papers say

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

“do not fully resolve token-level misalignment”
— GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
“However, these methods are still primarily based on token-level supervised objectives. Such objectives improve local next-token prediction, but speculative utility is inherently window-level and prefix-sensitive: once the accepted prefix is truncated due to an early mismatch, the remaining drafted tokens in the speculative window are invalidated.”
— Performance-Driven Policy Optimization for Speculative Decoding with Adaptive Windowing

Beaten on benchmarks

Head-to-head results where a newer method reports beating HASS. Values are copied from the source paper's tables — verify against the cited paper.

TriSpec beats HASS · Verification time (t_v) [Qwen3-HASS-MATH500]
0.070 vs 0.094
TriSpec: Ternary Speculative Decoding via Lightweight Proxy Verification
TriSpec beats HASS · Latency [Qwen3-HASS-MATH500]
83.03 vs 115.04
TriSpec: Ternary Speculative Decoding via Lightweight Proxy Verification
TriSpec beats HASS · Speed [Qwen3-HASS-MATH500]
45.26 vs 32.45
TriSpec: Ternary Speculative Decoding via Lightweight Proxy Verification
GRIFFIN beats HASS · SR [LLaMA2-7B, Temperature=0, MT-Bench]
3.12 vs 2.97
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · tau [LLaMA2-7B, Temperature=0, MT-Bench]
5.11 vs 4.97
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · SR [LLaMA2-7B, Temperature=0, Average]
3.28 vs 3.17
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · tau [LLaMA2-7B, Temperature=0, Average]
5.44 vs 5.26
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · SR [LLaMA3-8B, Temperature=0, MT-Bench]
3.09 vs 2.75
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · tau [LLaMA3-8B, Temperature=0, MT-Bench]
4.85 vs 4.63
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · SR [LLaMA3-8B, Temperature=0, Average]
3.35 vs 3.12
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · tau [LLaMA3-8B, Temperature=0, Average]
5.38 vs 5.13
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding
GRIFFIN beats HASS · SR [Vicuna-7B, Temperature=0, MT-Bench]
4.02 vs 3.91
GRIFFIN: Effective Token Alignment for Faster Speculative Decoding

Newer alternatives

Recent methods in the same sub-problem, not yet superseded in the knowledge base.