Research focus

AI for biology. Efficient intelligence.

I develop generative models for biological discovery and design, and low-precision methods that make large AI models more efficient to train and deploy.

I am a PhD student at the University of Maryland, advised by Heng Huang and working closely with Wei Chen. Previously, I received my B.S. in Biological Science from Tsinghua University.

Research interests

AI for biology
Generative modeling for virtual cells, mRNA design, and protein design. I develop models to predict cellular responses and generate biological sequences with desired functions.
Efficient machine learning
Post-training quantization, quantization-aware training, and efficient inference for language and multimodal models. I work on low-precision methods that reduce memory and compute costs while preserving model capabilities.
Long-term vision

My long-term mission is to build a future in which biological systems can be understood and responsibly programmed to improve human life, while intelligence is accessible to everyone. I pursue this vision through two complementary directions: AI for biology, to move from understanding life to designing it responsibly; and efficient machine learning, to pair stronger intelligence with substantially lower inference costs so advanced AI becomes a resource everyone can use.

Selected publications

Google Scholar

* Equal contribution.

  1. 2026

    MPQ: A Message-Passing View of Post-Training Quantization

    Yanshuo Chen, Chenxi Liu, Reza Shirkavand, and Heng Huang.

    NeurIPS 2026 · Poster.

    A message-passing approach to post-training quantization for large language models.

  2. 2026

    Grounding Multimodal Reasoning with Evidence-Ablated Negatives

    Chenxi Liu, Tianyi Xiong, Yanshuo Chen, and Heng Huang.

    NeurIPS 2026 · Poster.

    Uses targeted visual masking to generate hard negatives for reinforcement learning, reducing language-prior shortcuts and improving visually grounded multimodal reasoning.

  3. 2026

    Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following

    Tianyi Xiong, Yi Ge, Ming Li, Zuolong Zhang, Pranav Kulkarni, Kaishen Wang, Qi He, Zeying Zhu, Chenxi Liu, Ruibo Chen, Tong Zheng, Yanshuo Chen, Xiyao Wang, Renrui Zhang, Wenhu Chen, and Heng Huang.

    CVPR 2026 Highlight [paper] [project] [code]

    A benchmark for testing whether multimodal judges can follow diverse, fine-grained evaluation criteria.

  4. 2026

    Learning from Self-Debate: Preparing Reasoning Models for Multi-Agent Debate

    Chenxi Liu, Yanshuo Chen, Ruibo Chen, Tianyi Xiong, Tong Zheng, and Heng Huang.

    arXiv preprint. [paper] [code]

    Trains one reasoning model for both independent problem solving and productive multi-agent debate.

  5. 2025

    CodonTranslator: A Conditional Codon Language Model for Codon Optimization Across Life Domains

    Yanshuo Chen*, Yuming Zhang*, Joshua Li, Boxue Tian, and Heng Huang.

    bioRxiv preprint. [paper] [code] [model]

    A 150M-parameter conditional language model for cross-species codon optimization.

  6. 2025

    Fast and Scalable Wasserstein-1 Neural Optimal Transport Solver for Single-Cell Perturbation Prediction

    Yanshuo Chen, Zhengmian Hu, Wei Chen, and Heng Huang.

    Bioinformatics, ISMB/ECCB 2025. [paper] [arXiv] [code]

    A scalable optimal-transport method for predicting cellular responses to unseen perturbations.

  7. 2025

    Enhancing Privacy in Biosecurity with Watermarked Protein Design

    Yanshuo Chen*, Zhengmian Hu*, Yihan Wu, Ruibo Chen, Yongrui Jin, Marcus Zhan, Chengjin Xie, Wei Chen, and Heng Huang.

    Bioinformatics, 41(7). [paper] [code]

    Introduces statistical watermarking for protein design as a practical biosecurity safeguard.

  8. 2025

    A Watermark for Order-Agnostic Language Models

    Ruibo Chen*, Yihan Wu*, Yanshuo Chen, Chenxi Liu, Junfeng Guo, and Heng Huang.

    ICLR 2025. [paper] [arXiv]

  9. 2022

    Deep Autoencoder for Interpretable Tissue-Adaptive Deconvolution and Cell-Type-Specific Gene Analysis

    Yanshuo Chen*, Yixuan Wang*, Yuelong Chen, Yuqi Cheng, Yumeng Wei, Yunxiang Li, Jiuming Wang, Yingying Wei, Ting-Fung Chan, and Yu Li.

    Nature Communications, 13:6735. [paper] [code] [docs]

News

  • Two papers, MPQ and EAN, have been accepted as posters at NeurIPS 2026!
  • New blog post: Who Owns Intelligence?, on running AI on our own computers, what it costs, and who controls our data.
  • New blog post: running Qwen3.8-27B on a single RTX 5090, with measurements of quantization, long-context inference, and cache fidelity.
Earlier updates 5
  • Our Multi-Crit paper appeared at CVPR 2026 and was selected as a Highlight.
  • Recognized as an ICML 2026 Gold Reviewer.
  • Released Learning from Self-Debate.
  • Released CodonTranslator, its code, and pretrained model.
  • Our Wasserstein-1 neural optimal transport solver was published in Bioinformatics and presented at ISMB/ECCB.

Experience

  • Research Intern, Adobe Research Efficient Multi-Agent Workflow
    –

Teaching & mentoring

  • Co-Instructor, Selected Topics in Machine Learning (CMSC848G)
  • Mentor, High School Students Research Program
  • Mentor, REU research projects

Education & distinction

  • University of Maryland, College ParkPh.D. in Computer Science
  • Tsinghua UniversityB.S. in Biological Science
  • China National Biology OlympiadGold Medal