I am Tianyang Zhou (周天阳 in Chinese), a second-year PhD student in Information Systems and Management at Heinz College, Carnegie Mellon University, advised by Prof. Leman Akoglu.
My research focuses on the safety and interpretability of foundation models, especially large language models (LLMs) and LLM agents, with a current focus on LLM guardrails and on building reliable agents for complex real-world tasks.
Before CMU, I completed my Bachelor of Economics and Finance at Tsinghua University, with a minor in Statistics.
News
- Sep 2026 — Released a first-author preprint, When Less Compute Is More: Adaptive Early Exit Improves Pretrained Outlier Detection.
- May 2026 — Released a first-author preprint, Structured Prompt Optimization Meets Reinforcement Learning for Global and Local Interpretability over Complex Text.
- Mar 2026 — Co-authored a survey, Confidence as Control: A Survey of Confidence Utilization in Large Language Models.
- Aug 2025 — Started my PhD at the Heinz College, Carnegie Mellon University, advised by Prof. Leman Akoglu.
Publications
Preprints / Under Review
When Less Compute Is More: Adaptive Early Exit Improves Pretrained Outlier Detection
Preprint, 2026
Tianyang Zhou, Leman Akoglu.
Structured Prompt Optimization Meets Reinforcement Learning for Global and Local Interpretability over Complex Text
Preprint, 2026
Tianyang Zhou, Wenbo Chen, Pierre Jinghong Liang, Leman Akoglu.
Confidence as Control: A Survey of Confidence Utilization in Large Language Models
Preprint, 2026
Yubo Li, Tianyang Zhou, Xiaobin Shen, Yidi Miao, Rema Padman, Ramayya Krishnan.
Education
- Ph.D. in Information Systems and Management, Carnegie Mellon University — Heinz College, 2025 – present
- Advisor: Prof. Leman Akoglu
- Bachelor of Economics and Finance, Tsinghua University, 2021 – 2025
- Minor in Statistics
Contact
- Email: tzhou3@andrew.cmu.edu
- GitHub: @tyzhou42
- Google Scholar: Tianyang Zhou
- ORCID: 0009-0006-6122-7745
- LinkedIn: tianyang-zhou