Will Schwarzer

Hey there, I’m Will. I’m a PhD student at the University of Massachusetts Amherst, where I work with Philip Thomas, as well as Bruno Castro da Silva and Scott Niekum. My research goal is to ensure that the enormous promise of increasingly autonomous AI is realized both safely and efficiently. I’m currently working on projects in reward inference, adversarial robustness, robot predictability, Seldonian LLM development, Seldonian RL, and autonomous paper generation. I’m particularly interested in topics relating to the safe use of LLMs for autonomous machine learning research. Such topics include but are not limited to iterative reasoning tasks, tool/computer use, adversarial robustness in LLMs, and reward hacking.

Fun fact: this headshot was taken at an excellent studio in Beijing while I was studying Mandarin there. I’m always happy to chat in Chinese. Aside from languages, I also enjoy playing the piano and rock climbing. I particularly love exploring climbing gyms in new countries (tally so far: two in Beijing, one in Bangkok, one in Istanbul, one in Fort William, Scotland, and one in London).

Note to prospective undergraduate and master’s collaborators: please read this before reaching out.