Yue Zhao

Research Summary

My research builds reliable, robust, and scalable AI that advances science and benefits society. I focus on developing rigorous algorithmic foundations, advancing safety and interpretability in large models, and creating open systems that connect research with real-world impact.

1. Reliable AI Foundations: Detecting the Unexpected.
I develop fundamental algorithms and benchmarks for detecting rare, unseen, or abnormal patterns across modalities. This work unifies anomaly detection, out-of-distribution (OOD) detection, and automated model selection to ensure that AI systems remain reliable and predictable under uncertainty.
Keywords: Anomaly Detection, OOD Detection, Model Selection, Robust Learning

2. Trust & Safety in Large Language Models and Agents.
I study how to make large models and agentic systems safe, interpretable, and aligned under real-world conditions. My work investigates hallucination mitigation, privacy and security safeguards, jailbreak prevention, and dynamic evaluation frameworks for trustworthy reasoning and decision-making.
Keywords: LLM Safety, Hallucination Mitigation, Privacy & Security, Trust Evaluation

3. Foundation Models for Science & Society.
I apply foundation models and generative AI to scientific and societal domains, addressing challenges in climate forecasting, healthcare, and political or social decision-making. These efforts combine domain knowledge with foundation model reasoning to accelerate discovery and policy insights.
Keywords: AI for Science, Generative AI, Decision Modeling, Computational Social Science

4. Scalable, Automated & Open AI Systems.
I create efficient and reproducible machine learning systems that enable large-scale, open, and automated deployment of AI. My open-source work emphasizes distributed inference, workflow automation, and user-centric design, promoting transparent and accessible AI research for academia and industry alike.
Keywords: ML Systems, Automated ML, Open-source AI, Distributed Computing