man with black hair smiling.

Michael Li

I'm an incoming PhD student in CS at Northeastern University, advised by David Bau and Byron Wallace. I previously studied statistics and machine learning at Carnegie Mellon University.

I'm interested in developing our scientific understanding of neural networks, mostly by studying the underlying mechanisms and representations they learn. More broadly, I'm excited by foundational questions about phenomena in deep learning, e.g. generalization, emergent capabilities, scaling laws, and so on.

News

Jul 2026

How Much Do Circuits Tell Us? was accepted as an oral to the Sci-FM workshop at COLM 2026.

Apr 2026

Joining Northeastern University as a CS PhD student in Fall 2026!

Apr 2026

How Much Do Circuits Tell Us? won a best paper award (preliminary track) at the LTI Student Research Symposium.

Apr 2026

Model Internal Sleuthing was accepted to ACL 2026 (Main).

Oct 2025

Presented BERTology in the Modern World at the Interplay workshop at COLM 2025.

Publications

Diagram of circuit necessity, sufficiency, consistency, and specificity criteria.
How Much Do Circuits Tell Us? Measuring the Consistency and Specificity of Language Model Circuits
Michael Li, Nishant Subramani.
COLM Workshop on Scientific Understanding of Foundation Models (Oral), 2026.
Classifier pipeline extracting lexical identity and inflectional features from layer activations.
Model Internal Sleuthing: Finding Lexical Identity and Inflectional Features in Modern Language Models
Michael Li, Nishant Subramani.
ACL, 2026.
Probing accuracy heatmaps across layers for BERT, Qwen, and OLMo models.
BERTology in the Modern World
Michael Li, Nishant Subramani.
COLM Workshop on the Interplay of Model Behavior and Model Internals, 2025.
TCN model forecast of cumulative COVID-19 cases versus actual counts.
Predicting the Daily Counts of COVID-19 Infection Using Temporal Convolutional Networks
Michael Li, Fatemeh Esfahani, Li Xing, Xuekui Zhang.
Journal of Global Health, 2023.