Method Drift›Speculative decoding
Superseded baseline#84 of 151 most-superseded
AdaSpec
AdaSPEC: Selective Knowledge Distillation for Efficient Speculative DecodersSpeculative decoding · first seen Oct 22, 2025
superseded — cited as a baseline and beaten by newer methods
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
Newer alternatives
Recent methods in the same sub-problem, not yet superseded in the knowledge base.
- May 4, 2026
- component-aware self-speculative decodingComponent-Aware Self-Speculative Decoding in Hybrid Language ModelsMay 1, 2026
- Apr 22, 2026
- Apr 16, 2026
- Apr 2, 2026
- greedy multi-path block verification (GBV)Greedy Multi-Path Block Verification for Faster Decoding in Speculative SamplingFeb 18, 2026
- SDFPSDFP: Speculative Decoding with FIT-Pruned Models for Training-Free and Plug-and-Play LLM AccelerationFeb 5, 2026
- Feb 1, 2026
- CAS-Spec (Cascade Adaptive Self-Speculative Decoding)CAS-Spec: Cascade Adaptive Self-Speculative Decoding for On-the-Fly Lossless Inference Acceleration of LLMsOct 30, 2025
- Oct 26, 2025
- Oct 17, 2025
- Oct 1, 2025