YW Yuxiang Wei

PhD Student | Georgia Institute of Technology

Yuxiang Wei

I build foundation models for fMRI and benchmarks for reliable language agents.

At Georgia Tech, my research connects brain activity with language, vision, and multimodal neuroimaging through fMRI LLMs, diffusion-based stimulus reconstruction, and multimodal representation learning. Beyond neuroscience, I study how to evaluate LLM systems, especially long-term memory and coding agents.

Google Scholar LinkedIn GitHub

Research

Developing foundation models for brain signals and benchmarks for reliable AI agents.

Foundation Models

I build multimodal foundation models for fMRI that align brain activity with language and vision, supporting neural decoding, stimulus reconstruction, and cross-subject generalization.

Clinical Applications

I explore how foundation models and multimodal neuroimaging can support clinical problems such as early Alzheimer's disease assessment, with an emphasis on interpretable biomedical signals.

Agentic Memory

I study how long-term memory should be represented, organized, retrieved, and maintained in dialogue agents and AI clones, separating useful memory structure from unnecessary architectural complexity.

Benchmarks

I design benchmarks for LLM systems, including coding agents and memory systems, to understand how communication, context, and task design affect real-world performance.

Publications

Full Google Scholar profile

fMRI Representation and Foundation Models

Language-aligned and sequence-modeling approaches for general fMRI understanding, representation learning, and brain signal analysis.

  1. fMRI-LM: Towards a Universal Foundation Model for Language-Aligned fMRI Understanding. Yuxiang Wei, Yanteng Zhang, Xi Xiao, Chengxuan Qian, Tianyang Wang, Vince D. Calhoun. CVPR, 2026.
  2. Hierarchical Spatio-Temporal State-Space Modeling for fMRI Analysis. Yuxiang Wei, Anees Abrol, Vince D. Calhoun. RECOMB, 2025.

Generative Brain Decoding

Generative models that reconstruct visual stimuli from fMRI while improving interpretability and cross-subject generalization.

  1. MoRE-Brain: Routed Mixture of Experts for Interpretable and Generalizable Cross-Subject fMRI Visual Decoding. Yuxiang Wei, Yanteng Zhang, Xi Xiao, Tianyang Wang, Vince D. Calhoun. NeurIPS, 2026.

Long-Term Memory Benchmarks for LLM Agents

Evaluation frameworks for dialog memory and AI-clone memory that test whether agents can retrieve, maintain, and reason over long-horizon personal context.

  1. Does Memory Need Graphs? A Unified Framework and Empirical Analysis for Long-Term Dialog Memory. Sen Hu†, Yuxiang Wei†, JiaXin Ran, et al. ACL Main, 2026.
  2. CloneMem: Benchmarking Long-Term Memory for AI Clones. Sen Hu†, Zhiyu Zhang†, Yuxiang Wei†, Xueran Han, et al. ACL Main, 2026.

† Equal contribution.

Clinical Neuroimaging and Alzheimer's Disease

Multimodal neuroimaging and dynamic connectivity models for early Alzheimer's diagnosis, risk assessment, and interpretable clinical biomarkers.

  1. 4D Multimodal Co-attention Fusion Network with Latent Contrastive Alignment for Alzheimer's Diagnosis. Yuxiang Wei, Anees Abrol, Deqiang Qiu, James Lah, Allan I. Levey, Vince D. Calhoun. WACV, 2026.
  2. Spatiotemporal State Space Modeling of Dynamic Brain Connectivity in Cognitively Normal Individuals at Risk for Alzheimer's Disease. Yuxiang Wei, Anees Abrol, Deqiang Qiu, James Lah, Allan I. Levey, Vince D. Calhoun. ICASSP Oral, 2026.
  3. From Symptomatic to Pre-symptomatic: Adaptive Knowledge Distillation for Early Alzheimer's Detection. Yuxiang Wei, Anees Abrol, Vince D. Calhoun. IEEE Transactions on Biomedical Imaging, 2025.

General Deep Learning and Applied AI

Broader deep learning work across causal uplift modeling, visual recognition, prompt tuning, multimodal prediction, and domain-specific AI benchmarks.

  1. Scaling Vision Transformers for Functional MRI with Flat Maps. Connor Lane, Mihir Tripathy, Yuxiang Wei, Paul S. Scotti, et al. ICML, 2026.
  2. RoadBench: A Vision-Language Foundation Model and Benchmark for Road Damage Understanding. Xi Xiao, Yunbei Zhang, Janet Wang, Li Zhao, Yuxiang Wei, Tianyang Wang. WACV, 2026.
  3. Visual Instance-aware Prompt Tuning. Xi Xiao, Yunbei Zhang, Xingjian Li, Tianyang Wang, Xiao Wang, Yuxiang Wei, Jihun Hamm, Min Xu. ACM Multimedia, 2025.
  4. Multi-Treatment Multi-Task Uplift Modeling for Enhancing User Growth. Yuxiang Wei, Zhaoxin Qiu, Yingjie Li, Yuke Sun, Xiaoling Li. arXiv, 2024.
  5. Wavelet Integrated Attention Network with Multi-Resolution Frequency Learning for Mixed-Type Wafer Defect Recognition. Yuxiang Wei, Huan Wang. Engineering Applications of Artificial Intelligence, 2023.
  6. An Explainable Multi-View Deep Network for Prediction of Non-Imaging Phenotypes using Anatomical Multi-view Data. Yuxiang Wei, Yuqian Chen, Tengfei Xue, Fan Zhang, Lauren J. O'Donnell. MICCAI Workshop, 2023.
  7. Mixed-Type Wafer Defect Recognition with Multi-Scale Information Fusion Transformer. Yuxiang Wei, Huan Wang. IEEE Transactions on Semiconductor Manufacturing, 2022.
  8. Mixed-type Wafer Defect Pattern Recognition Framework based on Multi-faceted Dynamic Convolution. Yuxiang Wei, Huan Wang. IEEE Transactions on Instrumentation and Measurement, 2022.

Experience

Research and applied machine learning

2023 - Present

Graduate Research Assistant | TReNDS Center, Georgia Tech

Developed fMRI foundation and visual decoding models, along with state-space and knowledge-distillation methods for dynamic brain connectivity and early Alzheimer's risk analysis.

Summer 2026

Research Intern | Handshake AI

Host by: Curtis Northcutt

Building interactive coding benchmarks with user simulation.

Summer 2025

Applied Scientist Intern | Amazon

Host by: Richard Sun

Built graph-aware time-series forecasting models and spatiotemporal diffusion methods for probabilistic forecasts across large node graphs.

Summer 2024

Applied Scientist Intern | Tencent

Host by: Serlin Li

Designed multi-treatment, multi-task uplift modeling methods for user growth and robust uplift prediction under biased data collection.

2022 - 2023

Undergraduate Research Assistant | Golby Lab, Harvard Medical School

Host by: Lauren O'Donnell

Developed multi-view learning models for tabular diffusion MRI analysis, using white matter and gray matter features to predict brain-based phenotypes.

Spring 2022

Machine Learning Intern | HPC-AI Tech

Host by: You Yang

Contributed to ColossalAI, an open-source distributed deep learning system, by building Ray-based elastic training components and supporting Docker and Kubernetes deployment.

2020 - 2022

Undergraduate Research Assistant | Tsinghua University

Host by: Huan Wang

Developed transformer, convolutional, and autoencoder-based models for multi-label wafer defect recognition, including wavelet-based neural network components.

Awards

Travel Awards and Grants

  • CVPR Travel Award, 2026 ($600)
  • Georgia Tech SGA Conference Fund, 2025, 2026 ($1200)
  • RECOMB Student Travel Grant, 2025 ($1600)

Education

Training

PhD, Electrical and Computer Engineering

Georgia Institute of Technology | Expected May 2028

MS, Electrical and Computer Engineering

Georgia Institute of Technology | May 2025

BS, Electronic Information Engineering

University of Electronic Science and Technology of China | June 2023