Weidong Tang 汤维栋

Education

• 
  Xidian UniversitySep 2023 - Jul 2027 (expected)

  B.E., School of Telecommunication Engineering · GPA: 3.8/4.0

  Advised by Prof. Xiumei Wang.

Experience

• 
  Cardinal AI Lab, InfRecJul 2026 - Present

  Chief Research Officer (Founding Member)

  • Lead research on recursive self-improvement and long-horizon interaction, enabling agents to learn and control embodied robots.
  • Guide research from problem formulation to prototypes, working with engineering teams on real-world embodied AI applications.
• 
  AlibabaMay 2026 - Jul 2026

  MLLM Research Intern

  • Research: combined multimodal representations with OPD, exploring post-training, SFT, and industrial product benchmarks.
  • Business: studied image-text alignment data pipelines for product representations and retrieval in large-scale e-commerce.
• 
  TeleAIJan 2026 - Mar 2026

  LLM Research Intern

  • Focused on AI infrastructure for LLM training and inference on Huawei Ascend 910B, emphasizing efficient and stable execution.
  • Worked on precision alignment for LLM training and inference, addressing precision differences at the AI infrastructure level.
• 
  National University of Singapore · HPC-AI LabOct 2024 - Oct 2025

  Research Assistant

  Advised by Prof. Yang You, Prof. Wangbo Zhao, and Dr. Pengfei Zhou.

  • Focused on efficient AI and multimodal model evaluation, studying model efficiency and performance across research tasks.
  • Learned to lead research teams through collaborative projects, developing ideas and working toward research publications.

Publications

* Equal contribution · † Corresponding author · ‡ Project leader

1. 01
   The Trinity: Capability-Driven Recursive Self-Improvement for Multimodal Agents

   W. Tang^*, R. Wang^*, J. Zhang^*, H. Wang, H. Chen, Z. Zhang, Y. Ma, Z. Jiang, Z. Guo, H. Li, Z. Jiao, Z. Cui, X. Yin, S. Liang, Z. Wan, H. Lin, W. Zhao, K. Zhang, P. Zhou^†, O. Zhang^†, & Y. You^†.

   arXiv preprint, 2026 PaperGitHubProject Page
2. 02
   GroupToM-Bench: Benchmarking Group Theory of Mind and Nonlinear Social Emergence in MLLMs

   W. Tang, J. Li, Y. Hou, Z. Mei, C. Zhang, X. Wan, Z. Liang, P. Zhou^†, Y. You, & W. Zhao^†.

   ACL 2026 (19% acceptance rate, CCF-A top-tier conference) PaperDataset
3. 03
   MagicBench: Diagnosing Visual Agency Loss and Semantic Dependency in Multimodal LLMs

   T. D. Huang, W. Tang^‡, W. Q. Xu, & X. Guo^†.

   ACL 2026 (19% acceptance rate, CCF-A top-tier conference) PaperGitHub
4. 04
   Efficient Video Object Segmentation and Tracking with Recurrent Dynamic Submodel

   W. Tang, Z. Liang, X. Wan, C. Zhu, Z. Xu, P. Zhou, Y. Song, Y. You, & W. Zhao^†.

   CVPR 2026 (25% acceptance rate, CCF-A top-tier conference) Paper
5. 05
   Inference-Time Scaling for Visual AutoRegressive Modeling by Searching Representative Samples

   W. Tang, X. Wan, S. Li, & X. Wang^†.

   PRCV 2025 (29% acceptance rate, CCF-C) PaperGitHub
6. 06
   EarthVerse: Benchmarking Scientific Agents Across Dynamic Earth Systems and Natural Hazards

   Z. Cui, X. Yin, Y. Tang, X. Zhang, Y. Hu, S. Zhong, W. Tang, Y. Liang, W. Li, M. Jin, S. Pan, Y. Kang, D. Zhuang, & J. Zhao.

   arXiv preprint, 2026 PaperGitHubDatasetProject Page
7. 07
   IndustryBench-MIPU: Benchmarking Multi-Image Attribute Value Extraction for Industrial Products

   H. Qi, J. Cao, Y. Zhang, X. Wang, W. Tang, B. Chen, C. Huo, H. Pan, H. You, J. Li, Y. Wang, & L. Ding.

   arXiv preprint, 2026 PaperGitHubDataset
8. 08
   MDK12-Bench: A Multi-Discipline Benchmark for Evaluating Reasoning in Multimodal Large Language Models

   P. Zhou, X. Peng, F. Zhang, Z. Xu, J. Ai, Y. Qiu, W. Zhao, J. Song, C. Li, W. Tang, Z. Li, H. Zhang, Z. Li, X. Mao, Y. Feng, J. Sun, K. Wang, X. Chang, W. Shao, Y. You, & K. Zhang^†.

   AAAI 2026 (17% acceptance rate, CCF-A) PaperGitHub
9. 09
   EA-ViT: Efficient Adaptation for Elastic Vision Transformer

   C. Zhu, W. Zhao^†, H. Zhang, Y. Zhou, W. Tang, S. Wang, Z. Yuan, Y. Shang, X. Peng, K. Wang, & D. Yang^†.

   ICCV 2025 (25% acceptance rate, CCF-A top-tier conference) PaperGitHub
10. 10
    Pebench: A Fictitious Dataset to Benchmark Machine Unlearning for Multimodal Large Language Models

    Z. Xu, P. Zhou, W. Tang, J. Ai, W. Zhao, K. Wang, X. Peng, W. Shao, H. Yao, & K. Zhang^†.

    arXiv preprint, 2025 PaperDatasetProject Page
11. 11
    IEBench: Benchmarking Unlearning Identity-Event Entangled Concepts for MLLM-Based Systems in Agentic Edge Networks

    Z. Xu, P. Zhou, W. Tang, J. Ai, W. Zhao, K. Wang, W. Shao, X. Peng, S. Zhao, H. Yao, & K. Zhang^†.

    IEEE Transactions on Network Science and Engineering, 2026 (IF=7.9, JCR-Q1) Paper

Selected Projects

• WowPage

  A lightweight, open-source homepage template for researchers to showcase publications, projects, and academic profiles.