List your company

Mathematical Reasoning on GSM8K (MAT, Token/s)

6.77MAT (Accuracy)HedgeSpec
0.76922.32713.8855.4429Oct 22, 2025

Evaluation results

MethodLinks
HedgeSpecTarget Model=Llama-3.1-8B2025.106.7784.32
HedgeSpecTarget Model=Qwen-3-8B2025.106.673.59
HedgeSpecTarget Model=Qwen-3-32B2025.106.5742.37
EAGLETarget Model=Llama-3.1-8B2025.106.3878.32
UCBSpecTarget Model=Llama-3.1-8B2025.105.0566.11
EAGLETarget Model=Qwen-3-8B2025.105.0560.59
EXP3SpecTarget Model=Llama-3.1-8B2025.105.0363.28
UCBSpecTarget Model=Qwen-3-8B2025.105.0257.42
UCBSpecTarget Model=Qwen-3-32B2025.10537.5
EXP3SpecTarget Model=Qwen-3-8B2025.104.953.57
EXP3SpecTarget Model=Qwen-3-32B2025.104.8636.38
EAGLETarget Model=Qwen-3-32B2025.103.4424.86
Llama-3.1-8BTarget Model=Llama-3.1-8B2025.10118.26
Qwen-3-8BTarget Model=Qwen-3-8B2025.10114.82
Qwen-3-32BTarget Model=Qwen-3-32B2025.1018.24