List your company

Code Generation on SWE-bench OpenHands

3.7SpeedupHybrid Verified Decoding
0.76721.52862.293.0514May 31, 2026

Evaluation results

MethodLinks
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=Greedy Decoding2026.053.7
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=Greedy Decoding2026.053.31
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=SuffixDecoding2026.053.21
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=SuffixDecoding2026.052.6
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=Greedy Decoding2026.052.5
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=EAGLE32026.052.11
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=SuffixDecoding2026.051.52
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=Frequency-Based Hybrid Decoding2026.051.28
Hybrid Verified DecodingTarget Model=Llama3.1-8B, Reference Baseline=Frequency-Based Hybrid Decoding2026.051.26
Hybrid Verified DecodingTarget Model=Qwen3-8B, Reference Baseline=EAGLE32026.051.09
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=EAGLE32026.051.08
Hybrid Verified DecodingTarget Model=Qwen3-4B, Reference Baseline=Frequency-Based Hybrid Decoding2026.050.88