Hi, there! I’m Zihao (Gavin).

I study whether the explanations that large language models give for their answers, in particular chain-of-thought reasoning, faithfully report what drove those answers, and how to measure and train such faithfulness.

I build counterfactual and interpretability-based evaluations with proper baselines, and model organisms with known hidden influences as ground truth, toward trustworthy and auditable AI, especially in high-stakes domains. I am available for internships from May 2027 to August 2027.

I am a second-year CS PhD student at Northeastern University, advised by Professor Byron Wallace. I graduated cum laude from New York University with a Bachelor’s degree in Computer Science and Data Science. Previously, I worked in VisionX Lab as a research intern, advised by Professor Saining Xie. During my undergraduate program, I worked in OLAB and Memo Lab, advised by Professor Eric K. Oermann and Professor Shuo Chen respectively.

I have been fortunate to be advised by Professor Xie, Professor Oermann, and Professor Chen. I am also grateful for the mentorship from Lavender Y. Jiang, Xujin C. Liu, Professor Kyunghyun Cho, and Professor Samuel R. Bowman.

Email: yang.zih@northeastern.edu or gavin.yang@nyu.edu. CV: PDF.

Publications

Peer-reviewed

• Compared to What? Baselines and Metrics for Counterfactual Prompting

  Zihao Yang, Mosh Levy, Yoav Goldberg, Byron C. Wallace

  COLM 2026 arXiv Code Package

• Optimizing Packing and Shuffling Strategies for Enhanced Performance in Generative Language Models

  Yanbing Chen, Ruilin Wang, Zihao Yang, Lavender Jiang, Eric Oermann

  ACL SRW 2026 ACL Anthology

• Cambrian-S: Towards Spatial Supersensing in Video

  Shusheng Yang*, Jihan Yang*, Pinzhi Huang†, Ellis Brown†, Zihao Yang, Yue Yu, Shengbang Tong, Zihan Zheng, Yifan Xu, Muhan Wang, Rob Fergus, Yann LeCun, Li Fei-Fei, Saining Xie

  ICLR 2026 arXiv Project page Code Models Dataset Benchmark

  * equal contribution; † core contributor

• Medical large language models are vulnerable to data-poisoning attacks

  Daniel Alber, Zihao Yang, Anton Alyakin, …, Eric Oermann

  All 33 authors

  Daniel Alber, Zihao Yang, Anton Alyakin, Eunice Yang, Sumedha Rai, Aly Valliani, Jeff Zhang, Gabriel Rosenbaum, Ashley Amend-Thomas, David Kurland, Caroline Kremer, Alexander Eremiev, Bruck Negash, Daniel Wiggan, Michelle Nakatsuka, Karl Sangwon, Sean Neifert, Hammad Khan, Akshay Save, Adhith Palla, Eric Grin, Monika Hedman, Mustafa Nasir-Moin, Xujin Liu, Lavender Jiang, Michal Mankowski, Dorry Segev, Yindalon Aphinyanaphongs, Howard Riina, John Golfinos, Daniel Orringer, Douglas Kondziolka, Eric Oermann

  Nature Medicine 2025 Nature Medicine

• Intriguing Effect of the Correlation Prior on ICD-9 Code Assignment

  Zihao Yang, Chenkang Zhang, Muru Wu, Xujin Liu, Lavender Jiang, Kyunghyun Cho, Eric Oermann

  ACL SRW 2023 ACL Anthology Code

  

Preprints and manuscripts

• BPQA Dataset: Evaluating How Well Language Models Leverage Blood Pressures to Answer Biomedical Questions

  Chi Hang, Ruiqi Deng, Lavender Jiang, Zihao Yang, Anton Alyakin, Daniel Alber, Eric Oermann

  arXiv 2025 arXiv Dataset

• Language Models Can Guess Your Identities from De-identified Clinical Notes

  Lavender Jiang, Daniel Alber, Zihao Yang, Karl Sangwon, Xujin Liu, Kyunghyun Cho, Eric Oermann

  Under review OpenReview

Education

• 2025–present: Ph.D., Computer Science, Northeastern University, Boston, MA, USA
• 2020–2024: B.A., Computer Science & Data Science, Minor in Mathematics, New York University, New York, NY, USA