Haobo Xu 「徐浩博」

profile.png

Welcome to my homepage!

I am a second-year PhD student at the University of Illinois at Urbana-Champaign, advised by Prof. Hanghang Tong. Prior to this, I obtained my bachelor’s degree in Software Engineering from Tsinghua University.

My research interests include Efficient Machine Learning and (Multimodal) Large Language Models.

News

Sep 2026
Our new work “STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization” is just released. [Code/Paper] 📜
Aug 2026
Our two papers have been accepted to EMNLP 2026 (1 Main + 1 Findings)! 🎉
May 2026
Honored to receive the Gold Reviewer Award from ICML 2026!
Apr 2026
Our paper “Prune as You Generate: Online Rollout Pruning for Faster and Better RLVR” was accepted by ACL 2026 as an oral presentation! 🔥🔥
Sep 2025
Graduated from Tsinghua University and started my Ph.D. at UIUC! 🎓
Aug 2025
Our two papers were accepted by CoRL 2025! 🎉
Dec 2024
Grateful to be awarded the National Scholarship! 🎓
Nov 2024
Honored to be selected as a Top Reviewer at NeurIPS 2024!
Sep 2024
Our paper “DuQuant: Distributing Outliers via Dual Transformation Makes Stronger Quantized LLMs” was accepted by NeurIPS 2024 as an oral! 🎉🎉
May 2024
Our paper “SLOG: An Inductive Spectral Graph Neural Network Beyond Polynomial Filter” was accepted by ICML 2024! 🚀

Education

Experience

  • Research Scientist Intern @ Meta MRS Research Team

    May 2026 – Present
    Menlo Park, CA, USA

Selected Publications [full list]

* denotes equal contribution.

2026

  1. stepquant.png
    STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization
    Preprint, 2026
    Bingchen Yao*, Haobo Xu*, Haokun Lin, and 6 more authors
    Linear Attn Quantization Efficiency
  2. pill.png
    Predict, Don’t Iterate: Efficient Adaptive-Length Infilling for Diffusion Language Models
    EMNLP Main Conference, 2026
    Haobo Xu, Sirui Chen, Yuanchen Bei, and 5 more authors
    DLMs Generation Efficiency
  3. arrol.png
    Prune as You Generate: Online Rollout Pruning for Faster and Better RLVR
    ACL Main Conference (Oral Presentation), 2026
    Haobo Xu, Sirui Chen, Ruizhong Qiu, and 5 more authors
    Efficiency LLMs RLVR

2025

  1. tavla.png
    TA-VLA: Elucidating the Design Space of Torque-aware Vision-Language-Action Models
    CoRL, 2025
    Zongzheng Zhang*, Haobo Xu*, Zhuo Yang*, and 5 more authors
    Multi-modal Robotics
  2. qdlm.png
    Quantization Meets dLLMs: A Systematic Study of Post-training Quantization for Diffusion LLMs
    Machine Intelligence Research, 2025
    Haokun Lin*, Haobo Xu*, Yichen Wu, and 6 more authors
    Efficiency DLMs Quantization

2024

  1. duquant.png
    DuQuant: Distributing Outliers via Dual Transformation Makes Stronger Quantized LLMs
    NeurIPS (Oral Presentation, Acceptance Rate ~0.4%), 2024
    Haokun Lin*, Haobo Xu*, Yichen Wu*, and 6 more authors
    Efficiency LLMs Quantization

Awards & Honors

  • ICML Gold Reviewer Award, 2026
  • Excellent Graduate, Tsinghua University (清华大学优良毕业生), 2025
  • NeurIPS Top Reviewer Award, 2024
  • China National Scholarship (国家奖学金), 2024
  • NeurIPS Scholar Award, 2024
  • ICML Travel Award, 2024
  • Comprehensive Excellence Award, Tsinghua University, 2021–2024
  • First Provincial-level in Chinese Physics Olympiad, 2018, 2019
  • First Prize in National Olympiad in Informatics in Provinces (NOIP) of China, 2017, 2018

Services

  • Conference Reviewers:
    • IEEE BigData (2023), IJCNN (2026), AAAI (2024 - 2027), AISTATS (2025 - 2027), ICLR (2025 - 2027), ICML (2025 - 2026), NeurIPS (2024 - 2026), ACL Rolling Review (Oct. 2026).
  • Journal Reviewers:
    • ACM TIST, IEEE TSC, TMLR