Jiayuan Mao

News

• I'm recruiting PhD students this application cycle (Fall 2027). Please apply to the CIS PhD program at Penn by December 15th and list my name on your application. Unfortunately, I’m unable to respond to individual inquiries outside the official application process.
• To junior PhD/master/undergraduate students: if you would like to chat about life, career plan, or research ideas related to AI/ML, feel free to email me to schedule a meeting. I will dedicate 30 mins every week for such meetings. I encourage students from underrepresented groups to reach out and will prioritize these meetings.

  Underrepresented groups: including but not limited to gender/racial/ethnic minority groups.

• I'm co-organizing the Workshop on Space in Vision, Language, and Embodied AI and the Embodied Agent Interface Challenge at NeurIPS 2025.
• I'm co-organizing the Workshop on Memory and Vision at ICCV 2025.
• I will be giving a talk at the Workshop on Human-Robot-Scene Interaction and Collaboration at ICCV 2025.
• I'm co-organizing the Workshop on Resource-Rational Robot Learning at CoRL 2025.
• I'm co-organizing the Workshop on Programmatic Reinforcement Learning at RLC 2025.
• I'm co-organizing the Workshop on Programmatic Representations for Agent Learning at ICML 2025.
• I'm co-organizing the Workshop on Robot Planning in the Era of Foundation Models at RSS 2025.
• I'm co-organizing the Workshop on Visual Concepts and the Workshop on Foundation Models Meet Embodied Agents at CVPR 2025.
• I'm co-organizing the Workshop on Learning Meets Model-Based Methods for Contact-Rich Manipulation at ICRA 2025.
• I will be giving a talk at the Workshop on Foundation Models and Neuro-Symbolic AI for Robotics at ICRA 2025.
• I'm co-organizing the Workshop on Visual Concepts and the Workshop on Foundation Models for Embodied Agents at CVPR 2025.
• I'm co-organizing the "Learning Language through Grounding" tutorial and the "Foundation Models Meet Embodied Agents" tutorial at NAACL 2025.
• I'm co-organizing the Workshop on Planning in the Era of LLMs and the "Foundation Models Meet Embodied Agents" tutorial at AAAI 2025.
• I was selected as a Rising Star in EECS 2024.
• I was selected as a Rising Star in Generative AI 2024.

• I'm co-organizing the Workshop on Learning Effective Abstractions for Planning at CoRL 2024.
• I'm co-organizing and will be giving a talk at the Workshop on Visual Concepts at ECCV 2024.
• I will be giving a talk at the Bimanual Manipulation: On Kitchen Challenges workshop at ICRA 2024.
• I will be giving a talk at the Brown Robotics Talks at Brown University. (Slides: Compositional Action Representations)
• I will be giving a talk at the NSF Workshop on Hardware-Software Co-design for Neuro-Symbolic Computation.
• I will be giving a talk at the Manipulation Reading Group at the Robotics Institute at Carnegie Mellon University.
• I will be giving a talk at the Coordinated Science Laboratory Student Conference (CSLSC 2024) at the University of Illinois at Urbana-Champaign. (Slides: Integrated Learning and Planning)
• I will be giving a talk at the Robot Representations For Scene Understanding, Reasoning and Planning workshop at RSS 2023. (Slides: Neuro-Symbolic Concepts for Robotic Manipulation / Video)
• I'm co-organizing the Visually Grounded Interaction and Language (VIGIL) workshop at NAACL 2021.
• I'm co-organizing the Neuro-Symbolic Visual Reasoning and Program Synthesis tutorial at CVPR 2020.

PhD Students

• Julia (Wanjia) Fu (co-advised with Rachel Holladay)
• Junyu Zhang
• Yuchen Zhou

Teaching

• CIS 7000: Multi-Modal AI (Fall 2026)

Research Highlights

My research goal is to understand the principles that enable machines to continually acquire concepts from experience and use them to reason and plan in open-ended physical worlds. Scientifically, I study how the structure of tasks, the data recipe, and the inference algorithms jointly determine a machine’s ability to generalize beyond its training experience. Methodologically, I study deeply integrated neuro-symbolic systems. Their symbolic component captures the minimal structure needed for inference: the abstractions of the external environment, and the rules by which they can be manipulated through search, optimization, and probabilistic reasoning. Such structure can arise not only from language, but also from geometric and physical abstractions. The neural component grounds these abstractions in perception and physics, while end-to-end learning allows the representations and reasoning algorithms to adapt jointly from data.