ATHENA: Mathematical Reasoning with Thought Expansion

Benchmark Model Rank Results
math-word-problem-solving-on-asdiv-aATHENA (roberta-large)#1Execution Accuracy: 91
math-word-problem-solving-on-asdiv-aATHENA (roberta-base)#3Execution Accuracy: 86.4
math-word-problem-solving-on-math23kATHENA (roberta-large)#12Accuracy (training-test): 86.5
math-word-problem-solving-on-math23kATHENA (roberta-base)#13Accuracy (training-test): 84.4
math-word-problem-solving-on-mawpsATHENA (roberta-large)#3Accuracy (%): 93
math-word-problem-solving-on-mawpsATHENA (roberta-base)#6Accuracy (%): 92.2
math-word-problem-solving-on-svampATHENA (roberta-large)#16Execution Accuracy: 54.8
math-word-problem-solving-on-svampATHENA (roberta-base)#19Execution Accuracy: 45.6