REU Site: Recent Advances in Natural Language Processing

该项目的目标是让那些渴望在计算机科学领域，特别是人工智能和机器学习（AI-ML）领域从事高级职业的有动力的本科生在职业生涯早期获得积极的研究经验。该项目将在科罗拉多斯普林斯的科罗拉多大学为本科生开发一个夏季研究体验（REU）网站，每年吸引来自全国各地的10名参与者参与前沿研究，例如大型语言模型的使用和安全性，以及使用量子计算的新范式来更好地解决自然语言处理中的问题。 学生将编写软件来解决有趣和及时的问题，撰写论文发表，并将接触到申请研究生院的程序。该提案还将侧重于培训未来的计算机科学家，他们来自研究机会有限的机构、妇女和代表性不足的少数民族。研究经验将鼓励这些学生在未来的职业生涯中成为学术和非学术环境中的富有成效的研究人员。参与的学生将通过重点讲座，最新出版物的讨论和个性化阅读来介绍最新进展，如深度学习，大型语言模型和量子计算。特别是，学生将了解深度学习的最新创新，包括受变压器启发的架构，例如变压器双向编码器表示（BERT）和生成式预训练变压器（GPT）系列，这些架构在最近一段时间内为许多问题带来了持续改进的结果，这些问题多年来一直是研究的焦点。学生将对新问题进行实践研究，使用基于transformer的深度学习架构和量子计算方法进行实验，并通过书面论文和演示来传达他们的结果。示例项目将涉及控制大型语言模型的生成;保护语言模型免受攻击;以及使用量子计算解决诸如词性标记和命名实体识别等任务。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估而被认为值得支持。

The objective of this project is to expose motivated undergraduates who aspire to pursue advanced careers in Computer Science, in particular in Artificial Intelligence and Machine Learning (AI-ML), to active research experience early in their careers. This project will develop a summer Research Experience for Undergraduates (REU) site at the University of Colorado, Colorado Springs, to engage ten participants per year from around the country in cutting-edge research on topics such as the use and security of Large Language Models, and the use of the new paradigm of Quantum Computing to better solve problems in Natural Language Processing. The students will write software to solve interesting and timely problems, write papers for publication, and will be exposed to procedures for applying for graduate school. The proposal will also focus on training future computer scientists from institutions with limited research opportunities, women, and under-represented minorities. The research experience will encourage these students to be productive researchers in academic and non-academic environments during their future careers.The participating students will be introduced to the latest advances such as Deep Learning, Large Language Models, and Quantum Computing via focused lectures, discussion of latest publications, and personalized readings. In particular, the students will learn about latest innovations in Deep Learning, including Transformer-inspired architectures, such as the Bidirectional Encoder Representations from Transformers (BERT) and the Generative Pre-Trained Transformers (GPT) series, that have ushered in a recent period of sustained improved results to many problems that have been the focus of research for years. The students will perform hands-on research on novel problems, conduct experiments using Transformer-based Deep Learning architectures and Quantum Computing methods, and communicate their results through written papers and presentations. Example projects will involve controlling generation of large language models; defending language models against attacks; and the use of Quantum Computing in solving tasks such as part-of-speech tagging and named entity recognition.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.