SubmissionNumber#=%=#222
FinalPaperTitle#=%=#Calc-CMU at SemEval-2024 Task 7: Pre-Calc - Learning to Use the Calculator Improves Numeracy in Language Models
ShortPaperTitle#=%=#
NumberOfPages#=%=#8
CopyrightSigned#=%=#Vishruth Veerendranath
JobTitle#==#
Organization#==#Carnegie Mellon University
5000 Forbes Avenue, Pittsburgh, PA
Abstract#==#Quantitative and numerical comprehension in language is an important task in many fields like education and finance, but still remains a challenging task for language models. While tool and calculator usage has shown to be helpful to improve mathematical reasoning in large pretrained decoder-only language models, this remains unexplored for smaller language models with encoders. In this paper, we propose Pre-Calc, a simple pre-finetuning objective of learning to use the calculator for both encoder-only and encoder-decoder architectures, formulated as a discriminative and generative task respectively. We pre-train BERT and RoBERTa for discriminative calculator use and Flan-T5 for generative calculator use on the MAWPS, SVAMP, and AsDiv-A datasets, which improves performance on downstream tasks that require numerical understanding. Our code and data are available at https://github.com/calc-cmu/pre-calc.
Author{1}{Firstname}#=%=#Vishruth
Author{1}{Lastname}#=%=#Veerendranath
Author{1}{Username}#=%=#vishruthnath
Author{1}{Email}#=%=#vveerend@andrew.cmu.edu
Author{1}{Affiliation}#=%=#Carnegie Mellon University
Author{2}{Firstname}#=%=#Vishwa
Author{2}{Lastname}#=%=#Shah
Author{2}{Username}#=%=#vishwa_shah_27
Author{2}{Email}#=%=#vishwavs@andrew.cmu.edu
Author{2}{Affiliation}#=%=#Carnegie Mellon University
Author{3}{Firstname}#=%=#Kshitish
Author{3}{Lastname}#=%=#Ghate
Author{3}{Username}#=%=#kghate
Author{3}{Email}#=%=#kghate@andrew.cmu.edu
Author{3}{Affiliation}#=%=#Carnegie Mellon University

==========
èéáğö