Deep learning techniques for math word problems
This paper presents a comprehensive investigation into Curriculum Learning (CL) applied to Math Word Problem (MWP) solving, examining its efficacy across a spectrum of scales and difficulty levels. Our study encompasses extensive experiments conducted on both small-scale and large-scale language...
Saved in:
| 主要作者: | |
|---|---|
| 其他作者: | |
| 格式: | Final Year Project |
| 語言: | English |
| 出版: |
Nanyang Technological University
2024
|
| 主題: | |
| 在線閱讀: | https://hdl.handle.net/10356/178278 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 機構: | Nanyang Technological University |
| 語言: | English |
| 總結: | This paper presents a comprehensive investigation into Curriculum Learning (CL)
applied to Math Word Problem (MWP) solving, examining its efficacy across a spectrum
of scales and difficulty levels. Our study encompasses extensive experiments conducted
on both small-scale and large-scale language models across three distinct MWP datasets,
featuring problems of varying difficulty ranges. Through rigorous evaluation, we find
that curriculum learning yield better performance than traditional training in MWP
solving tasks. We also find the potential of anti-curriculum learning in solving hard
mathematical questions. Additionally, we offer an in-depth analysis of the mechanisms
and effects of curriculum learning and its variations in MWP solving |
|---|