Research overview

My research is in Natural Language Processing and Machine Learning, with an emphasis on applications in health.

Working in the domain of health naturally motivates the methodological problems that I have worked on. For example, these include: model interpretability; learning with limited supervision from diverse sources; human-in-the-loop/hybrid systems; and trustworthiness of model outputs. For more details, see recent publications here.

On the applications side, one thread of my research concerns developing language technologies to automate (or semi-automate) biomedical evidence synthesis. Here is an episode of the NLP highlights podcast in which I discuss this work, here is a (brief) talk I gave at SciNLP 2020, and here is an article written for a lay audience about the effort. Elsewhere, I have worked on models for processing notes in Electronic Health Records.

A random sample of recentish publications

Jaden Fiotto-Kaufman, Alexander R. Loftus, Eric Todd, Jannik Brinkmann, Koyena Pal, Dmitrii Troitskii, Michael Ripa, Adam Belfki, Can Rager, Caden Juang, Aaron Mueller, Samuel Marks, Arnab Sen Sharma, Francesca Lucchetti, Nikhil Prakash, Carla Brodley, Arjun Guha, Jonathan Bell, Byron C. Wallace and David Bau. NNsight and NDIF: Democratizing Access to Foundation Model Internals ICLR; 2025.

Alberto Mario Ceballos-Arroyo, Monica Munnangi, Jiuding Sun, Karen Zhang, Jered McInerney, Byron C. Wallace and Silvio Amir. Open (Clinical) LLMs are Sensitive to Instruction Phrasings BioNLP; 2024.

Somin Wadhwa, Chantal Shaib, Silvio Amir and Byron C. Wallace. Who Taught You That? Tracing Teachers in Model Distillation ACL (Findings); 2025.

News

09/19/2025 Open Philanthropy grant

Open Philanthropy has awarded me a grant to work on "Mechanistic Interpretability for Healthcare"

09/18/2025 NeurIPs spotlight

Our paper, "Learning the Wrong Lessons: Syntactic-Domain Spurious Correlations in Language Models", received a Spotlight designation at NeurIPs 2025 (top ~4% of 21575)

01/16/2024 ICLR Spotlight

Our paper, Evaluating the Zero-shot Robustness of Instruction-tuned Language Models was accepted as Spotlight (top 5%) at ICLR 2024

10/01/2022 Helping radiologists navigate EHR

We have received a new R01 from the NIH/NLM to work on neural summarization methods to aid diagnosis (collaboration with Dr. Geoffrey Young at Brigham and Women's Hospital).

Support

My work has been supported with grants from the National Institutes of Health, National Science Foundation (including a CAREER grant), the Army Research Office, Seton hospital, Amazon and seed funds from Brown University.