List your company

Multi-hop Question Answering on HotpotQA (Speedup)

3.77Speedup FactorHybrid Verified Decoding
0.69161.49082.293.0892May 31, 2026

Evaluation results

MethodLinks
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=SuffixDecoding2026.053.77
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=SuffixDecoding2026.052.7
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=Greedy Decoding2026.052.57
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=Greedy Decoding2026.052.23
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=Greedy Decoding2026.052.03
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=SuffixDecoding2026.051.97
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=Frequency-Based Hybrid Decoding2026.051.37
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=Frequency-Based Hybrid Decoding2026.051.02
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=Frequency-Based Hybrid Decoding2026.050.96
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=EAGLE32026.050.95
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=EAGLE32026.050.93
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=EAGLE32026.050.81