List your company

Mathematical Word Problem Solving on GSM8K (sampled 500 test)

95AccuracySelf-consistency
91.8892.6993.594.31Oct 2, 2025

Evaluation results

MethodLinks
Self-consistencyBackend Model=gpt-4o2025.1095
Chain-of-Thought (CoT)Backend Model=gpt-4o2025.1094
Self-askBackend Model=gpt-4o2025.1094
FOR-PromptingBackend Model=gpt-4o, Host=True2025.1094
FOR-PromptingBackend Model=gpt-4o, Host=False2025.1094
Single-promptBackend Model=gpt-4o2025.1092