List your company

Multi-path speculative decoding on Qwen held-out (test)

1.31Throughput Ratio ImprovementNaiveTree NDE
0.97721.06361.151.2364Feb 19, 2026

Evaluation results

MethodLinks
NaiveTree NDEVerification algorithm=NaiveTree, Neural Dynamic Expansion=true2026.021.31
NSS NDEVerification algorithm=NSS, Neural Dynamic Expansion=true2026.021.2
SpecInfer NDEVerification algorithm=SpecInfer, Neural Dynamic Expansion=true2026.021.15
SpecTr NDEVerification algorithm=SpecTr, Neural Dynamic Expansion=true2026.021.14
Khisti NDEVerification algorithm=Khisti, Neural Dynamic Expansion=true2026.020.99