Yue Zhao
Avatar of Yue Zhao
Assistant Professor
Thomas Lord Department of Computer Science
School of Advanced Computing

University of Southern California

Los Angeles, CA, USA
Email:

Collaboration with Me. I am open to external opportunities for invited talks, research collaborations, and employment (only on the part-time/advising/visiting basis). Let us have a chat by email. I frequently visit major cities, e.g., Seattle, NYC, Chicago, Boston, Atlanta, and Bay Area to meet people, give talks, and host social events.

Research Interests: My research focuses on creating reliable, efficient, and accessible AI systems. These efforts span from building theoretical foundations to delivering practical tools and real-world impact:

  1. Reliable AI Foundations: Detecting the Unexpected. I develop core algorithms to identify anomalies, outliers, and out-of-distribution (OOD) instances, enhancing AI systems' resilience to unfamiliar or rare events across diverse settings.
    Keywords: Anomaly Detection, Outlier Mining, OOD Detection, Robust AI
  2. Graph Intelligence: Learning from Relational Structures. Extending foundational detection methods, I build graph-based algorithms to model and analyze complex relationships, enabling discovery in domains such as finance, science, and online platforms.
    Keywords: Graph Learning, Graph Outlier Detection, Open-set Learning, Structured Data Mining
  3. Foundation Models in Practice: From Insights to Impact. By combining robust modeling with the expressiveness of large language models (LLMs) and generative AI, I explore applications in critical fields like scientific discovery and societal forecasting.
    Keywords: LLMs, Generative AI, AI for Science, AI for Society, Drug Discovery
  4. Scalable, Open, and Automated AI Systems. To amplify the reach of these methods, I build scalable, reproducible, and open-source ML systems—such as PyOD (25M+ downloads)—empowering global adoption across academia and industry.
    Keywords: Automated ML, Open-source AI, ML Systems, Distributed Computing

Biography.

Lab Openings. We are warmly welcoming new members to the FORTIS Lab!

Ph.D. Students (no more than 1 Ph.D. student for Fall 2026):
  • Due to the large number of interested candidates, future Ph.D. students are ideally have multiple published, relevant papers (not necessarily with me) in top-tier ML, System, CV, or NLP conferences/journals.
Research Interns (Any Time, All Year Round):
  • We welcome both undergraduate and graduate interns from USC and other institutions.
  • Preferred candidates are located in North America for time zone compatibility.
  • I do not have any summer interns -- I am also enjoying summer for fun :)
Application Process: To apply for either opportunities, complete the Application Form, email me after submitting the form, and review the FORTIS Lab website for more information before reaching out.

✈ News and Travel

[Mar 2025] We have a new paper on hierarchical cross-modal alignment for decoupled multimodal representation learning (DecAlign); see our Preprint!

[Mar 2025] We have a new paper exploring a causal approach to mitigating hallucinations in Vision-Language Models (VLMs); see our Preprint!

[Mar 2025] We have a new paper on secure and efficient on-device OOD detection without backpropagation (SecDOOD); see our Preprint!

[Mar 2025] Join the newly established ACM Transactions on AI for Science (TAIS) as an Associate Editor!

[Mar 2025] We have a new paper, TRUSTEVAL: A Dynamic Evaluation Toolkit on Trustworthiness of Generative Foundation Models, accepted to NAACL 2025 Demo Track; see our Preprint soon!

[Feb 2025] We have a new paper, DPU: Dynamic Prototype Updating for Multimodal Out-of-Distribution Detection, accepted to CVPR 2025; see our Preprint!

[Feb 2025] We have a new paper, Edit Away and My Face Will Not Stay: Personal Biometric Defense against Malicious Generative Editing, accepted to CVPR 2025; see our Preprint!

[Feb 2025] We have a new paper on multimodal LLMs for time series anomaly detection (Can Multimodal LLMs Perform Time Series Anomaly Detection?); see our Preprint!

[Feb 2025] We have a new paper on model extraction attacks and defenses in distributed computing environments; see our Preprint!

[Feb 2025] We have a new paper on the trustworthiness of generative foundation models ("On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective"); see our Preprint and Project Website!

[Feb 2025] We have a new paper, “ClimateLLM,” proposing a frequency-aware foundation model for efficient and accurate global weather forecasting. See our Preprint!

[Feb 2025] We have a new survey paper on LLM-based Active Learning, covering selection, generation, and its impact on modern AI pipelines. See our Preprint!

[Jan 2025] We have a new paper, MetaOOD: Automatic Selection of OOD Detection Models, accepted to ICLR 2025 and the KDD Workshop on Resource-Efficient Learning for Knowledge Discovery (Best Paper Award); see our Preprint!

[Jan 2025] We have a new paper, PyOD 2: A Python Library for Outlier Detection with LLM-powered Model Selection, accepted to The Web Conference 2025 Demo Track; see our Preprint!

🏅 Awards and Grants

As Principal Investigator (August 2023 onwards)
Prior to Principal Investigator Role (Before August 2023)