List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM8K (Sample Size Sensitivity)

128Samples RequiredAdaptive SC
1.1234.066799.94Nov 15, 2025
Updated 4mo ago

Evaluation results

MethodLinks
Adaptive SCSC@n=128, Model=Llama-3B2025.111282.13
Fixed-AllocationSC@n=128, Model=Llama-3B2025.11343.4
Blend-ASCSC@n=128, Model=Llama-3B2025.11224.29
Fixed-AllocationSC@n=64, Model=Llama-3B2025.11154.09
Fixed-AllocationSC@n=128, Model=Qwen-Math2025.11143.4
Adaptive SCSC@n=64, Model=Llama-3B2025.11134.36
Blend-ASCSC@n=64, Model=Llama-3B2025.11115.33
Adaptive SCSC@n=128, Model=Qwen-Math2025.11112.13
Fixed-AllocationSC@n=64, Model=Qwen-Math2025.1194.09
Blend-ASCSC@n=128, Model=Qwen-Math2025.1194.29
Adaptive SCSC@n=64, Model=Qwen-Math2025.1164.36
Blend-ASCSC@n=64, Model=Qwen-Math2025.1165.33