About Me
I am an incoming first-year CS masters student at the University of Maryland, College Park. I received my BS in Statistics and Machine Learning at Carnegie Mellon University where I graduated with University Honors. My current research experiences lie in LLM agents however I hope to explore other domains throughout my masters to further hone in on my research interests. In my undergraduate years I have been fortunate to work with several distinguished advisors on cutting-edge projects.
Under the guidance of Prof. Karthik Narasimhan’s at the Princeton NLP Group, I led research on PersonaGym, the first evaluation framework for persona agents in Large Language Models. With Prof. Cornelia Caragea, I developed ImplicitAVE, the first open-sourced dataset for implicit attribution value extraction, which was accepted to ACL Findings 2024. I worked with Prof. Daphne Ippolito on LLM output control to control for resonse length of instruction-tuned models. Additionally, I collaborated with Yijia Zhou and Prof. Diyi Yang on a framework for evaluating Human-AI collaboration between users and Language Model agents across multiple tasks.
I am always looking to collaborate on projects and can be reached at vsamuel@umd.edu
News
- May 19, 2025: Excited to release CIE: Controlling Language Model Text Generations Using Continuous Signals!
- Dec 20, 2024: Excited to release Collaborative Gym: A Framework for Enabling and Evaluating Human-Agent Collaboration!
- Nov 29, 2024: Our paper titled Towards Data Contamination Detection for Modern Large Language Models: Limitations, Inconsistencies, and Oracle Challenges was accepted into COLING 2025.
- Sept 18, 2024: Excited to release Towards Data Contamination Detection for Modern Large Language Models: Limitations, Inconsistencies, and Oracle Challenges!
- July 28, 2024: Excited to release PersonaGym!
- July 8, 2024: Our paper titled Can LLMs Augment Low-Resource Reading Comprehension Datasets? Opportunities and Challenges was accepted into ACL Student Research Workshop 2024.
- May 15, 2024: Our paper titled ImplicitAVE: An Open-Source Dataset and Multimodal LLMs Benchmark for Implicit Attribute Value Extraction was accepted into ACL Findings 2024