DDD (Speculative decoding): superseded — cited as a baseline and beaten by newer methods. 0 paper(s) critique it, 4 beat it on benchmarks — #20 of 151 most-superseded. Sub-problem: cluster led by EAGLE-3. Newer alternatives in the same sub-problem include D^2SD, TreeFlash, Hybrid Verified Decoding, Bastion, Draft-OPD.

Superseded baseline#20 of 151 most-superseded

DDD

Speculative decoding

superseded — cited as a baseline and beaten by newer methods

0 papers critique it · 4 beat it on benchmarks

Beaten on benchmarks

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

ECHO beats DDD · Avg. Speedup [Vicuna-13B]
5.25 vs 3.97
ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios
ECHO beats DDD · Avg. Speedup [LLaMA3.1-8B]
4.30 vs 3.83
ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios
ECHO beats DDD · Avg. Speedup [LLaMA3.3-70B]
5.35 vs 4.43
ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios
ECHO beats DDD · Avg. Speedup [Qwen3-8B]
2.74 vs 2.28
ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios
ECHO beats DDD · Avg. Speedup [Qwen3-32B]
2.37 vs 2.07
ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios
ECHO beats DDD · Avg. Speedup [Qwen3-235B]
2.23 vs 1.77
ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios
EVICT beats DDD · Average [Temperature = 0]
202.82 vs 165.70
Making Every Verified Token Count: Adaptive Verification for MoE Speculative Decoding
EVICT beats DDD · Average [Temperature = 1]
195.17 vs 160.99
Making Every Verified Token Count: Adaptive Verification for MoE Speculative Decoding
LDLP-DDD beats DDD · Tok/s [MT-Bench]
66.86 vs 66.33
AdaEAGLE: Optimizing Speculative Decoding via Explicit Modeling of Adaptive Draft Structures
LDLP-DDD beats DDD · Tok/s [Alpaca]
66.16 vs 65.41
AdaEAGLE: Optimizing Speculative Decoding via Explicit Modeling of Adaptive Draft Structures
LDLP-DDD beats DDD · Tok/s [HumanEval]
79.07 vs 77.71
AdaEAGLE: Optimizing Speculative Decoding via Explicit Modeling of Adaptive Draft Structures
LDLP-DDD beats DDD · Tok/s [GSM8K]
71.20 vs 69.39
AdaEAGLE: Optimizing Speculative Decoding via Explicit Modeling of Adaptive Draft Structures

Newer alternatives

Recent methods in the same sub-problem, not yet superseded in the knowledge base.