Yuning Yang

AI Researcher at FDA

I am a full-time AI researcher at FDA CDER AI review team advised by Qi Liu, Hadi Kharrazi, and David C. Fajgenbaum. My work primarily focuses on building internal AI review tools at the FDA to ensure that AI can reliably and efficiently support reviewers in their day-to-day evaluation of various cases. Outside of work, I also collaborate with academic researchers and professors, including Yuzhe Yang and Ying Nian Wu as well. My recent research interests are in Agentic AI in Healthcare and Foundation model in Pharmacology.

Before joining the FDA, I earned an M.Eng. in Artificial Intelligence from the University of California, Los Angeles, and a B.Sc. in Computer Science and Quantitative Economics at the University of California, Irvine. I was also fortunate to be advised by Amir M. Rahmani at the UCI HealthSciTech Group, and Guang Cheng at UCLA, where I researched synthetic data through deep generative modeling.

If you are a student interested in research about AI for health or would like to collaborate, please email me!

Contact: yuningyang at ucla dot edu / Yuning dot Yang at fda dot hhs dot gov

Selected Publications [full list]

(*) denotes equal contribution

ACL Main

Quantifying Preferences of Vision-Language Models via Value Decomposition in Social Media Contexts

Jingxuan Li*, Yuning Yang*, Shengqi Yang, Yizhou Zhao, and Ying Nian Wu

In The 63rd Annual Meeting of the Association for Computational Linguistics, 2025.

Abstract PDF Code Website

The recent progress in Vision-Language Models (VLMs) has broadened the scope of multimodal applications. However, evaluations often remain limited to functional tasks, neglecting abstract dimensions such as personality traits and human values. To address this gap, we introduce Value-Spectrum, a novel Visual Question Answering (VQA) benchmark aimed at assessing VLMs based on Schwartz’s value dimensions that capture core human values guiding people’s preferences and actions. We design a VLM agent pipeline to simulate video browsing and construct a vector database comprising over 50,000 short videos from TikTok, YouTube Shorts, and Instagram Reels. These videos span multiple months and cover diverse topics, including family, health, hobbies, society, technology, etc. Benchmarking on Value-Spectrum highlights notable variations in how VLMs handle value-oriented content. Beyond identifying VLMs’ intrinsic preferences, we also explore the ability of VLM agents to adopt specific personas when explicitly prompted, revealing insights into the adaptability of the model in role-playing scenarios. These findings highlight the potential of Value-Spectrum as a comprehensive evaluation set for tracking VLM preferences in value-based tasks and abilities to simulate diverse personas. The complete code and data are available at https://github.com/Jeremyyny/Value-Spectrum.
Under Review

TimeAutoDiff: Generation of Heterogeneous Time Series Data via Latent Diffusion Model

Namjoon Suh, Yuning Yang, Din-Yin Hsieh, Qitong Luan, Shirong Xu, Shixiang Zhu, and Guang Cheng

In Submission.

PDF Code