List your company

Speculative Decoding on MedQA

6.47Match Rate (MAT)HedgeSpec
0.78122.25813.7355.2119Oct 22, 2025

Evaluation results

MethodLinks
HedgeSpecBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.106.4777.26
HedgeSpecBackbone=Qwen-3-8B, Inference Setting=Greedy2025.106.0865.82
HedgeSpecTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.106.0172.65
HedgeSpecBackbone=Qwen-3-32B, Inference Setting=Greedy2025.105.9238.6
HedgeSpecTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.105.8463.67
HedgeSpecTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.105.4639.29
EagleBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.104.9666.48
Vanilla EagleTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.104.6558.82
UCBSpecBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.104.4661.14
EXP3SpecBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.104.2558.13
UCBSpecBackbone=Qwen-3-8B, Inference Setting=Greedy2025.104.2548.96
UCBSpecBackbone=Qwen-3-32B, Inference Setting=Greedy2025.104.1930.99
UCBSpecTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.104.1241.97
UCBSpecTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.104.1153.79
EXP3SpecTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.104.1152.23
EXP3SpecBackbone=Qwen-3-8B, Inference Setting=Greedy2025.104.0745.02
UCBSpecTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.104.0629.09
EXP3SpecBackbone=Qwen-3-32B, Inference Setting=Greedy2025.103.9628.76
EXP3SpecTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.103.9540.96
EXP3SpecTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.103.9227.6
EagleBackbone=Qwen-3-8B, Inference Setting=Greedy2025.103.8444.13
Vanilla EagleTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.103.7642.89
EagleBackbone=Qwen-3-32B, Inference Setting=Greedy2025.102.5718.33
Vanilla EagleTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.102.5518.68
LLaMA-3.1-8B-ITTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.10118.81
Qwen-3-8BTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.10114.7
Qwen-3-32BTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.1018.38
LLaMA-3.1-8B-ITBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.10118.57
Qwen-3-8BBackbone=Qwen-3-8B, Inference Setting=Greedy2025.10114.59
Qwen-3-32BBackbone=Qwen-3-32B, Inference Setting=Greedy2025.1018.2