Haoyang Chen

Research Assistant @ UMN APEX Lab

selfie_portrait.jpg

I’m Haoyang Chen, a Research Assistant at the UMN APEX Lab, advised by Prof. Caiwen Ding. I received my B.S. in Computer Science from the University of Minnesota, Twin Cities in May 2026. I also work closely with Shiyang Li and Yuebo Luo.

My research aims to turn World Action Models (WAMs) into robot policies that work in the real world. This requires three coupled properties, which I approach through algorithm-system co-design:

  • Long-horizon consistency: persistent memory and state, so a WAM stays consistent with what it has already seen and done over long tasks.
  • Efficient inference: fast enough for closed-loop control on real robots. I contribute to vLLM-Omni on WAM inference optimization.
  • Robustness: policies that hold up under the variability of real environments and user demands, beyond curated benchmarks.

I’m applying to PhD programs for Fall 2027.

news

Sep 2026 Our paper CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging is accepted at NeurIPS 2026 (Evaluations and Datasets Track) as a poster!
May 2026 Graduated from the University of Minnesota, Twin Cities with a B.S. in Computer Science.
Feb 2026 Joined the UMN APEX Lab, working with Prof. Caiwen Ding and Shiyang Li on MLsys-related research.

selected publications

  1. CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging
    Shiyang Li*, Haoyang Chen*, Mattia Fazzini, and Caiwen Ding
    In Advances in Neural Information Processing Systems (NeurIPS), Evaluations and Datasets Track, 2026
  2. Skills in Weights, Memory in Code: Hybrid Learning for Memory-Dependent Robot Manipulation
    Yunhao Zhao*, Zhenyang Ni*, Haoyang Chen, Ruohan Zhang, and Qi Zhu
    arXiv preprint arXiv:2608.09410, 2026