I am a tenure-track assistant professor in the Department of Computer Science at the University of Maryland, College Park (UMD). I am a member of the CLIP (Computational Linguistics and Information Processing) lab.
My expertise is at the intersection of natural language processing (NLP), empirical machine learning, and explainable AI. I research the interpretability of language models (LMs) and other neural networks, with the goal of increasing their transparency, reliability, safety, and utility. I also focus on providing natural language explanations to users of LMs that are faithful to models’ internal operations.
Previously, I was a postdoctoral researcher at the Allen Institute for AI (Ai2) and the University of Washington, working with Ashish Sabharwal and Hannaneh Hajishirzi. Before that, I received my M.S. and Ph.D. degrees in computer science from Georgia Tech, where I was advised by Mark Riedl.
Short Bio for Talks
Recent Updates
- Fall 2026: Thanks to AI2, AIM, and Martian for recent funding supporting the lab.
- Fall 2026: One paper at COLM on interpreting latent reasoning models.
- Fall 2026: One paper at Interspeech on interpreting audio LMs.
- Fall 2026: The Actionable Interpretability workshop will be happening again at COLM 2026! Submit by June
21st24th. - Summer 2026: We will be at NEMI!
- Summer 2026: Recent preprints on contrastive linear directions predictive of emergent misalignment and finding circuits that drive cognitive anchoring behavior in LLMs.
- Summer 2026: Congrats to Nishkal, Swastik, and Navita on their paper on predicting steering vector success from geometric properties accepted to the Mechanistic Interpretability workshop at ICML. Poster. Full paper coming soon!
- Summer 2026: I'll be at ICML; also on a panel at KAIST on LM Reliability & Safety.
- Summer 2026: I'll be speaking at the Columbia ML Summer School.
- Summer 2026: New ICML position paper on "Actionable Interpretability" that spun out of our ICML 2025 workshop.
- Summer 2026: Congrats to Shramay for having a paper accepted to ACL.
- Summer 2026: Congrats to Chenlong, Yuhang, Zhihan, and Dongping on their CVPR Findings paper (and my first paper at a vision conference).
- Spring 2026: Thanks to Coefficient Giving for supporting the lab's work on chain-of-thought monitoring. Press Release
- Spring 2026: New preprint, led by Stephen, on tracing steering vector circuits for steering long-form generation.
- Spring 2026: Congrats to Connor for having a paper on interpreting latent reasoning models at the ICLR Latent & Implicit Reasoning workshop.
- Spring 2026: A couple of recent news articles about our research: UMD CS Department, UMIACS
- Spring 2026: Congrats to Kartik for having a paper accepted to EACL.
- Fall 2025: I'm now a member of the NSF National Deep Inference Fabric's advisory board.
- Fall 2025: (Remote) invited talks at UCLA and Cambridge.
- Fall 2025: Invited talk at the Interplay workshop at COLM. See you in Montreal!
- Fall 2025: Starting as an assistant professor in the CS department at the University of Maryland. Go terps! press release.
- Summer 2025: Congrats to Alec Bunn on having a paper accepted to the ACL 2025 GEM workshop.
- Summer 2025: Invited talk at University of British Columbia.
- Summer 2025: Our Mechanistic Interpretability Benchmark (MIB) was accepted to ICML 2025. Check out the website for all the benchmark resources. I am also organizing the Actionable Interpretability Workshop at ICML. See you in Vancouver!
- Spring 2025: Two papers accepted to ICLR 2025, the former as a spotlight.
- Winter 2024: Selected as a Rising Star in Machine Learning.
- Winter 2024: Recognized as an outstanding area chair at EMNLP 2024.
- Winter 2024: Paper proposing a taxonomy for model noncompliance accepted at NeurIPS 2024 Datasets and Benchmarks.
- Winter 2024: Attending EMNLP in Miami. Checkout our 2 Findings papers and our position paper on mechanistic interpretability at the BlackBoxNLP workshop. Slides from my talk at BlackBoxNLP: here.
- Fall 2024: Attending COLM in Philly.
- Fall 2024: Selected as a Rising Star in Generative AI.
- Summer 2024: Check out our NAACL 2024 tutorial on Explanation in the Era of Large Language Models. See you in Mexico City!
- Spring 2024: One paper, The Unreasonable Effectiveness of Easy Training Data for Hard Tasks, will appear at ACL 2024.
- Spring 2024: Gave a guest lecture in the graduate-level LLMs course at Washington University in St. Louis.
- Winter 2023: Recording of Sarthak Jain and I's invited talk ("Is Attention = Explanation? Past, Present, and Future") at the Big Picture workshop is available here (from 1:57).
- Winter 2023: Gave a talk to the Washington State Senate's Environment, Energy, and Technology Committee on "What is AI?".
- Winter 2023: Recognized as a top reviewer at NeurIPS 2023.
- Winter 2023: Two papers at EMNLP 2023, and giving an invited talk at the Big Picture workshop with Sarthak Jain. See you in Singapore!
- Winter 2023: Selected as a Rising Star in EECS.
- Winter 2023: Self-Refine published at NeurIPS.
- Fall 2023: Talks at UC Irvine, UCSD, and USC.
- Summer 2023: Slides and recording (from 7:45) of my invited talk at the ACL Natural Language Reasoning and Structured Explanations workshop are available.
- Summer 2023: Awarded an Outstanding Area Chair award (top 1.5% of reviewers and chairs) at ACL 2023.
- Spring 2023: Quoted in this article about language model interpretability.
- Fall 2022: Talks at various NLP groups at the University of Washington (Tsvetshop, H2Lab, and Treehouse).
- Fall 2022: Two Findings papers at EMNLP 2022.
- Fall 2022: Co-organizing the BlackBoxNLP workshop at EMNLP 2022. See you in Abu Dhabi!
We are grateful to the following sponsors for their support of our research:
