Cumulative Reasoning with Large Language Models
Open paper
Benchmark
Model
Rank
Results
math-word-problem-solving-on-math
CR (GPT-4-turbo model, w/ code)
#26
Accuracy: 72.2
math-word-problem-solving-on-math
CR (GPT-4 model, w/o code)
#60
Accuracy: 58.0
Rank counts only results with a code link.