Bingxin Xu

CS PhD Student @ USC

BingxinXu_LinkedIn.jpg
xubingxi@usc.edu

Hi! I’m a second-year Ph.D. student in Computer Science at University of Southern California, advised by Prof. Emilio Ferrara.

Before joining USC, I spent five years as a Data Scientist in industry. My academic background includes an M.S. in Data Science and a B.S. in Computer Science from the Illinois Institute of Technology, where I also served as a research assistant under the mentorship of Prof. Yan Yan.

My research focuses on Efficient and Trustworthy AI, spanning from Large Language Model (LLM), Vision-Language Model (VLM), to Vision-Language-Action Model (VLA).

News

[2026.05] Start my research intern at Samsung Research America for 2026 summer.

[2025.04] One first author paper is accepted to ACL 2026 Findings.

[2024.06] One co-first author paper is accepted to ICCV 2025.

Selected Publications

For a full list of publications, please refer to this page.

2026

  1. SilentDrift: Exploiting Action Chunking for Stealthy Backdoor Attacks on Vision-Language-Action Models
    SilentDrift: Exploiting Action Chunking for Stealthy Backdoor Attacks on Vision-Language-Action Models
    Bingxin Xu,  Yuzhang Shang,  Binghui Wang,  and Emilio Ferrara
    ACL Findings 2026 PDF
    2026

2025

  1. LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models
    LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models
    Bingxin Xu*,  Yuzhang Shang*,  Mu Cai*,  Yong Jae Lee,  and Yan Yan
    ICCV 2025 GitHub stars PDF
    2025
  2. ButterflyQuant: Ultra-low-bit LLM Quantization through Learnable Orthogonal Butterfly Transforms
    ButterflyQuant: Ultra-low-bit LLM Quantization through Learnable Orthogonal Butterfly Transforms
    Bingxin Xu,  Zhen Dong,  Oussama Elachqar,  and Yuzhang Shang
    arXiv 202509 PDF
    2025

2024

  1. freePruner: A Training-free Approach for Large Multimodal Model Acceleration
    freePruner: A Training-free Approach for Large Multimodal Model Acceleration
    Bingxin Xu,  Yuzhang Shang,  Yunhao Ge,  Qian Lou,  and Yan Yan
    arXiv 202411 PDF
    2024
  2. Interpolating Video-LLMs: Toward Longer-sequence LMMs in a Training-free Manner
    Interpolating Video-LLMs: Toward Longer-sequence LMMs in a Training-free Manner
    Bingxin Xu*,  Yuzhang Shang*,  Weitai Kang,  Mu Cai,  Yuheng Li,  Zehao Wen,  Zhen Dong,  Kurt Keutzer,  Yong Jae Lee,  and Yan Yan
    arXiv 202409 PDF
    2024

(*) denotes equal contribution

Awards

USC Annenberg Fellowship Research Symposium Award
2026/04
USC Annenberg Fellowship
2025/08 - Present
USC Viterbi School of Engineering Fellowship
2025/08 - Present
Cyrus Tang Scholarship Award
2023 - 2024

Education

Ph.D. Student in Computer Science
○ University of Southern California
○ Advisor: Prof. Emilio Ferrara
2025/08 - Present
M.S. in Data Science & B.S. in Computer Science
○ Illinois Institute of Technology
2017 - 2020

Work Experience

Research Intern
○ Samsung Research America
2026/05 - 2026/08
Research Assistant
○ CVM Lab at Illinois Institute of Technology
2023/01 - 2025/05
Data Scientist
○ Discover Financial Services, Project44
2020/06 - 2025/08

Services

Conference reviewer of ICLR 2025, ICML 2026