Dan Schumacher

I am an NLP researcher at the University of Texas at San Antonio, advised by Dr. Anthony Rios. My work explores the boundaries of LLM capability such as how models handle numerical context. It also includes human-centered studies and mechanistic interpretability.

I come from an unconventional background that heavily informs my approach to research. I began my professional life as a classical guitarist and instructor, eventually transitioning into public service at the San Antonio Central Library. While there, I managed procurement for Banned Books Week and experienced the real-world impact of information access—most notably helping a disabled patron call his mother for the first time in two years.

Driven by the desire to build better information systems, I taught myself to code while working two jobs. This led me to an M.S. in Data Analytics, where I met my advisor, Dr. Anthony Rios, who convinced me to pursue a Ph.D. and join the UTSA NLP Lab.

Research Focus

My current work centers on Context-Specific Reasoning, specifically probing how LLMs process varied data types like time series and temporal data. I am equally interested in Mechanistic Interpretability—understanding the internal “why” behind model behavior—and Human-AI Collaboration, where I study how users (particularly students) perceive and identify LLM agents in live, collaborative settings.

Selected Publications

Offline

When I’m not at the Human-Centered NLP Lab, I’m usually:

• Training and competing in dog agility with my wife and our Border Collies.
• In the boxing ring.
• Reading Sci-Fi novels.

Musical Background

Before my transition into research, I spent my time as a professional musician. Here are a few performances and projects: