Solving Challenging Math Word Problems Using GPT-4 Code Interpreter with Code-based Self-Verification

Benchmark Model Rank Results
math-word-problem-solving-on-mathGPT-4-code model (CSV, w/ code, SC, k=16)#12Accuracy: 84.3
math-word-problem-solving-on-mathGPT-4-code model (CSV, w/ code)#24Accuracy: 73.5
math-word-problem-solving-on-mathLogicNet (with code interpreter)#30Accuracy: 71.2
math-word-problem-solving-on-mathGPT-4-code model (w/ code)#35Accuracy: 69.7
math-word-problem-solving-on-mathGPT-4-code model (w/o code)#51Accuracy: 60.8