SubmissionNumber#=%=#277
FinalPaperTitle#=%=#Bit_numeval at SemEval-2024 Task 7: Enhance Numerical Sensitivity and Reasoning Completeness for Quantitative Understanding
ShortPaperTitle#=%=#
NumberOfPages#=%=#12
CopyrightSigned#=%=#Xinyue Liang
JobTitle#==#
Organization#==#
Abstract#==#In this paper, we describe the methods used for Quantitative Natural Language Inference (QNLI), and Quantitative Question Answering (QQA) in task1 of Semeval2024 NumEval.  The challenge's focus is to enhance the model's quantitative understanding consequently improving its performance on certain tasks. We accomplish this task from two perspectives: (1) By integrating real-world numerical comparison data during the supervised fine-tuning (SFT) phase, we enhanced the model's numerical sensitivity. (2) We develop an innovative reward model scoring mechanism, leveraging reinforcement learning from human feedback (RLHF) techniques to improve the model's reasoning completeness.
Author{1}{Firstname}#=%=#Xinyue
Author{1}{Lastname}#=%=#Liang
Author{1}{Username}#=%=#xyliang
Author{1}{Email}#=%=#luna.liang000@gmail.com
Author{1}{Affiliation}#=%=#Beijing Institute of Technology
Author{2}{Firstname}#=%=#Jiawei
Author{2}{Lastname}#=%=#Li
Author{2}{Username}#=%=#jwli22
Author{2}{Email}#=%=#jarvi_lee@163.com
Author{2}{Affiliation}#=%=#Beijing Institute of Technology
Author{3}{Firstname}#=%=#Yizhe
Author{3}{Lastname}#=%=#Yang
Author{3}{Username}#=%=#yizheyang
Author{3}{Email}#=%=#youngizzet@gmail.com
Author{3}{Affiliation}#=%=#Beijing Institute of Technology
Author{4}{Firstname}#=%=#Yang
Author{4}{Lastname}#=%=#Gao
Author{4}{Username}#=%=#gyang
Author{4}{Email}#=%=#gyang@bit.edu.cn
Author{4}{Affiliation}#=%=#Beijing Institute of Technology

==========
èéáğö