Xianyu Chen
Menu

Xianyu Chen

Portrait of Xianyu Chen

My research makes multimodal AI reason and explain more like people, by learning from where humans look.

AI Research Scientist at Fasikl, a NeuroAI company, where I develop deep learning for peripheral neural signals. Ph.D. in Computer Science, University of Minnesota (2025), with papers at ECCV (Oral), CVPR and IEEE TPAMI.

CV (PDF) Résumé (PDF) Email
Google Scholar GitHub

2026 Distinguished Dissertation Award, University of Minnesota · one of two university-wide

About the dissertation

Research

All publications

My dissertation, Building Human-Like Machine Intelligence, uses human visual attention (where people look, and in what order) in three steps: model it, align AI attention with it, and explain it.

  • Photo of a woman holding a vase with the question "Is the vase the same color as the scarf?" and two scanpaths: one behind the correct answer "no", one behind an incorrect answer.

    Modeling

    CVPR 2021

    Scanpaths in VQA

    The first model of human scanpaths in visual question answering, trained with reinforcement learning on whole gaze sequences.

  • A beach photo and a roundabout photo, each shown three times with a different observer's scanpath of numbered fixation circles.

    Modeling

    CVPR 2024

    Beyond Average

    Predicts how each individual viewer, not the average one, scans an image.

  • The problem "How to Involve a Pet in Christmas" with two illustrated steps: an ornament with the pet's picture, and a photo collage.

    Alignment

    CVPR 2022 · Co-firstTPAMI 2024

    VisualHow

    20K+ real-life problems with illustrated step-by-step solutions; supervising model attention with human attention improves solution prediction.

  • Street photo with a scanpath of circled fixations joined by arrows; speech bubbles explain two fixations: "There is a person standing on the sidewalk" and "There is a person walking on the sidewalk".

    Explainability

    ECCV 2024Oral · top 2.3%

    GazeXplain

    Predicts where people look and describes, in words, what each fixation looks at.

Experience

Full CV (PDF)

  1. Oct 2025 – Present

    Fasikl Inc. · Research Scientist – AI (intern 2023–2025)

    Co-author of SARM (IEEE TBME 2026); co-inventor on two patent applications for closed-loop stimulation control.

  2. Sep 2019 – Dec 2025

    University of Minnesota · Ph.D. in Computer Science

    Advisor: Prof. Catherine Qi Zhao. Doctoral research 2019–2024.

  3. Summer 2019

    Tencent AI Lab · Research Intern

  4. Aug – Dec 2018

    Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences · Research Intern

    Few-shot object detection (IEEE TIP 2020, AAAI 2020).

  5. Sep 2011 – Jun 2018

    Sun Yat-sen University · M.E. and B.E.

    Statistical signal processing for wireless and visible light communication (IEEE TSP 2017, IEEE WCL 2019).

Honors

  • 2026

    Distinguished Dissertation Award, University of Minnesota; University nominee for the CGS/ProQuest Distinguished Dissertation Award

  • 2025, 2026

    Outstanding Reviewer, WACV 2025 and CVPR 2026

  • 2019, 2020

    College of Science and Engineering Graduate Fellowship, University of Minnesota

  • 2013, 2017

    National Scholarship, Ministry of Education of China

Service · Reviewer for CVPR, ICCV, ECCV, AAAI, WACV and BMVC, and for IEEE TPAMI, TIP, TNNLS, TMM, TCSVT and ACM TOMM.

Contact

Email
Profiles
Short bio
For talks and introductions

Xianyu Chen is an AI Research Scientist at Fasikl. He received his Ph.D. in Computer Science from the University of Minnesota in 2025, advised by Prof. Catherine Qi Zhao. His research uses human visual attention to build multimodal AI that reasons and explains more like people, with papers at CVPR, ECCV (Oral), IJCAI and IEEE TPAMI. His dissertation received the University of Minnesota's 2026 Distinguished Dissertation Award.