LLMs as a resource for language-independent essay grading
Subrahmanya Aditya Vasanth Devarakonda, & Dr. Cynthia Matuszek, Department of Computer Science and Electrical Engineering.
In this work, we explore whether judicious use of frontier models can allow fair evaluation of written work despite errors introduced by non-native language use, focusing on the concepts and ideas presented rather than execution.
Grounded Language Learning by Robots
Dr. Cynthia Matuszek, Department of Computer Science and Electrical Engineering.
The Interactive Robotics and Language lab is a research laboratory in UMBC’s Computer Science and Electrical Engineering department. We study robotics and natural language processing, with the goal of bringing the fields together: developing robots that everyday people can talk to, telling them to do tasks or about the world around them. This approach to learning to understand language in the physical space that people and robots occupy is called grounded language acquisition. Our goal is to build robots that can perform tasks in noisy, real-world environments, instead of being pre-emptively programmed to handle a fixed set of predetermined tasks. The aim of our research program is to construct robots that interact with people seamlessly in unanticipated, dynamic environments and problem spaces. We focus on statistical learning approaches that let robots learn about the world entirely from multi-modal interactions with end users.
Accelerating K-Means++
Edward Raff, Department of Computer Science and Electrical Engineering
Frank Ferraro, Department of Computer Science and Electrical Engineering
Cynthia Matuszek, Department of Computer Science and Electrical Engineering
K-Means++, and its distributed variant K-Means++, have become de facto tools for selecting the initial seeds of k-means. Over the past decade since their introduction, no uniformly superior algorithms have been developed. Instead, we focus on accelerating these already well known methods. While retaining the exact same results, we can get 10-100x speedups by using the triangle inequality to selectively avoid redundant checks.