List your company

Question Answering on NaturalQA (Accuracy, Rejection Rate, Speedup)

38AccuracyVIA-SD
24.4827.9931.535.01Jun 10, 2026

Evaluation results

MethodLinks
VIA-SDModel=Qwen 7B→72B2026.0638322.7
VIA-SDModel=Qwen 7B→72B, Method=VIA-SD (ours), Speedup=2.70×2026.063832—
Faster CascadesModel=Qwen 7B→72B2026.0637412.15
Faster CascadesModel=Qwen 7B→72B, Method=Faster Cascades, Speedup=2.15×2026.063741—
Speculative DecodingModel=Qwen 7B→72B2026.0636471.55
BiLDModel=Qwen 7B→72B2026.0636431.95
Cascade SDModel=Qwen 7B→72B2026.0636451.7
SWiFTModel=Qwen 7B→72B2026.0636422.05
CLaSPModel=Qwen 7B→72B2026.0636402.25
Speculative DecodingModel=Qwen 7B→72B, Method=Speculative Decoding, Speedup=1.55×2026.063647—
Cascade SDModel=Qwen 7B→72B, Method=Cascade SD, Speedup=1.70×2026.063645—
VIA-SDModel=Gemma2 2B→27B2026.0634302.61
VIA-SDModel=Gemma2 2B→27B, Method=VIA-SD (ours), Speedup=2.61×2026.063430—
Faster CascadesModel=Gemma2 2B→27B2026.0633402.1
VIA-SDModel=LLaMA2 7B→70B2026.0633332.55
VIA-SDModel=Qwen 7B→14B2026.0633202.25
Faster CascadesModel=Gemma2 2B→27B, Method=Faster Cascades, Speedup=2.10×2026.063340—
VIA-SDModel=LLaMA2 7B→70B, Method=VIA-SD (ours), Speedup=2.55×2026.063333—
VIA-SDModel=Qwen 7B→14B, Method=VIA-SD (ours), Speedup=2.25×2026.063320—
Speculative DecodingModel=Gemma2 2B→27B2026.0632451.54
BiLDModel=Gemma2 2B→27B2026.0632411.92
Cascade SDModel=Gemma2 2B→27B2026.0632421.73
SWiFTModel=Gemma2 2B→27B2026.0632411.93
CLaSPModel=Gemma2 2B→27B2026.0632392.21
Faster CascadesModel=LLaMA2 7B→70B2026.0632422
Faster CascadesModel=Qwen 7B→14B2026.0632261.8
Speculative DecodingModel=Gemma2 2B→27B, Method=Speculative Decoding, Speedup=1.54×2026.063245—
Cascade SDModel=Gemma2 2B→27B, Method=Cascade SD, Speedup=1.73×2026.063242—
Faster CascadesModel=LLaMA2 7B→70B, Method=Faster Cascades, Speedup=2.00×2026.063242—
Faster CascadesModel=Qwen 7B→14B, Method=Faster Cascades, Speedup=1.80×2026.063226—
Speculative DecodingModel=LLaMA2 7B→70B2026.0631481.45
BiLDModel=LLaMA2 7B→70B2026.0631441.85
Cascade SDModel=LLaMA2 7B→70B2026.0631461.65
SWiFTModel=LLaMA2 7B→70B2026.0631431.9
CLaSPModel=LLaMA2 7B→70B2026.0631412.1
Speculative DecodingModel=Qwen 7B→14B2026.0631321.35
BiLDModel=Qwen 7B→14B2026.0631291.65
Cascade SDModel=Qwen 7B→14B2026.0631301.5
SWiFTModel=Qwen 7B→14B2026.0631281.75
CLaSPModel=Qwen 7B→14B2026.0631271.85
Speculative DecodingModel=LLaMA2 7B→70B, Method=Speculative Decoding, Speedup=1.45×2026.063148—
Cascade SDModel=LLaMA2 7B→70B, Method=Cascade SD, Speedup=1.65×2026.063146—
Speculative DecodingModel=Qwen 7B→14B, Method=Speculative Decoding, Speedup=1.35×2026.063132—
Cascade SDModel=Qwen 7B→14B, Method=Cascade SD, Speedup=1.50×2026.063130—
VIA-SDModel=Gemma2 2B→9B2026.0628172.1
VIA-SDModel=Gemma2 2B→9B, Method=VIA-SD (ours), Speedup=2.10×2026.062817—
Faster CascadesModel=Gemma2 2B→9B2026.0627221.75
VIA-SDModel=LLaMA2 7B→13B2026.0627222.05
Faster CascadesModel=Gemma2 2B→9B, Method=Faster Cascades, Speedup=1.75×2026.062722—
VIA-SDModel=LLaMA2 7B→13B, Method=VIA-SD (ours), Speedup=2.05×2026.062722—
Speculative DecodingModel=Gemma2 2B→9B2026.0626271.32
BiLDModel=Gemma2 2B→9B2026.0626241.67
Cascade SDModel=Gemma2 2B→9B2026.0626261.43
SWiFTModel=Gemma2 2B→9B2026.0626251.61
CLaSPModel=Gemma2 2B→9B2026.0626241.67
Faster CascadesModel=LLaMA2 7B→13B2026.0626281.68
Speculative DecodingModel=Gemma2 2B→9B, Method=Speculative Decoding, Speedup=1.32×2026.062627—
Cascade SDModel=Gemma2 2B→9B, Method=Cascade SD, Speedup=1.43×2026.062626—
Faster CascadesModel=LLaMA2 7B→13B, Method=Faster Cascades, Speedup=1.68×2026.062628—
Speculative DecodingModel=LLaMA2 7B→13B2026.0625341.28
BiLDModel=LLaMA2 7B→13B2026.0625311.55
Cascade SDModel=LLaMA2 7B→13B2026.0625321.4
SWiFTModel=LLaMA2 7B→13B2026.0625301.6
CLaSPModel=LLaMA2 7B→13B2026.0625291.7
Speculative DecodingModel=LLaMA2 7B→13B, Method=Speculative Decoding, Speedup=1.28×2026.062534—
Cascade SDModel=LLaMA2 7B→13B, Method=Cascade SD, Speedup=1.40×2026.062532—