Qihua Dong's

selfie_lowres.jpg

I am currently a PhD at Northeastern University, Boston. I graduated from City University of Hong Kong with a major in computer science and a minor in math.

My research interests focus on reasoning and visual understanding in (M)LLMs, including reinforcement learning and tool-use agents. My prior experience spans multimodal LLMs, image segmentation, and medical image analysis.

ps: You may reach me by email or GitHub. Welcome to collaborate!

News

Aug 2026 Our Ref-Adv-s benchmark is now integrated into EvalScope (ModelScope), with support for standardized evaluation and report generation. Check out the evaluation guide!
May 2026 Excited to join Meta SuperIntelligence Lab as a Research Scientist Intern, working on multimodal LLM training!
Apr 2026 Two papers accepted at ACL 2026: a Findings paper on a hierarchical visual agent with compact visual/text context for chart reasoning, and a Main Conference survey on thinking with images.
Mar 2026 Our Amazon work, Visual Reasoning through Tool-supervised Reinforcement Learning, was accepted to CVPR 2026 Findings and is now available on arXiv. If you find it interesting, feel free to upvote it on Hugging Face Papers.
Mar 2026 The code and data for our ICLR 2026 paper Ref-Adv: Exploring MLLM Visual Reasoning in Referring Expression Tasks are released at ref-adv.github.io.

Projects

The authors with * contributed equally to the work
  1. Visual Reasoning through Tool-supervised Reinforcement Learning
    Qihua Dong, Gozde Sahin, Pei Wang, Zhaowei Cai, and 3 more authors
    CVPR 2026 Findings, 2026
    Internship Work
    Paper
  2. Ref-Adv: Exploring MLLM Visual Reasoning in Referring Expression Tasks
    Qihua Dong, Yang Kuo, Ju Lin, Handong Zhao, and 5 more authors
    Proc. ICLR, 2026
    Ref-Adv-s is now integrated into EvalScope (ModelScope).
    Paper Code
  3. CoT Referring: Chain-of-Thought Grounding Improve Localization in Referring Expression Tasks
    Qihua Dong, Luis Figueroa, Handong Zhao, Kushal Kafle, and 4 more authors
    arXiv preprint, 2025
    Internship Work
    Paper
  4. Volume-Preserving Registration for 3D Medical Images with Tumors
    Qihua Dong*, Hao Du*, Ying Song, Yan Xu, and 1 more author
    Proc. ICCV, 2023
    Paper Code
  5. Weakly-Supervised Camouflaged Object Detection with Scribble Annotations
    Ruozhen He*, Qihua Dong*, Jiaying Lin, and Rynson W. H. Lau
    Proc. AAAI, 2023
    Paper Code
  6. Weakly-Supervised 3D Medical Image Segmentation using Geometric Prior and Contrastive Similarity
    Hao Du*, Qihua Dong*, Yan Xu, and Jing Liao
    IEEE Transactions on Medical Imaging, 2023
    Paper