List your company

Math Word Problem Reasoning on ASDiv (AUROC)

0.6818AUROCTOKENSAR + CoT-UQ
0.4765040.5298020.58310.636398Feb 24, 2025

Evaluation results

MethodLinks
TOKENSAR + CoT-UQStrategy=AP, Backbone=Mistral 7B2025.020.6818
Probas-mean + CoT-UQStrategy=AP, Backbone=Mistral 7B2025.020.674
TOKENSARStrategy=AP, Backbone=Mistral 7B2025.020.5843
Probas-meanStrategy=AP, Backbone=Mistral 7B2025.020.5726
P(True) + CoT-UQStrategy=SE, Backbone=Mistral 7B2025.020.513
P(True)Strategy=SE, Backbone=Mistral 7B2025.020.4844