CSI-Agent Group Embodied Intelligence at SKKU
CSI-Agent Group at Sungkyunkwan University – where our embodied AI loops move from simulation to real robots.
Frame
Robot Manipulation
Practice runs with our 6-DoF platform in the robotics bay.
News
Three Papers Accepted to NeurIPS 2025(2 Spotlight)
- NeSyPr: Neurosymbolic Proceduralization For Efficient Embodied Reasoning
- Policy Compatible Skill Incremental Learning via Lazy Learning Interface
- Towards Reliable Code-as-Policies: A Neuro-Symbolic Framework for Embodied Task Planning
We are excited to announce that our CSI-Agent Group at SKKU has had three papers accepted to NeurIPS 2025, with two of them receiving the prestigious Spotlight recognition. Out of 21,575 submissions, only 688 (≈3%) were selected as Spotlights, making this a particularly meaningful achievement.
These papers, led and co-authored by our talented students, focus on embodied and physical AI, highlighting the lab’s growing impact in advancing neurosymbolic reasoning, incremental learning, and reliable embodied task planning.
Two Papers Accepted to ICLR 2025
- NeSyC: A Neuro-symbolic Continual Learner
- Model Risk-sensitive Offline RL
- Advancing embodied AI and safe learning
Our innovative work on Neuro-symbolic Continual Learning (NeSyC) and Model Risk-sensitive Offline Reinforcement Learning has been accepted to ICLR 2025, one of the premier venues for deep learning research.
Research Areas
Agentic AI research on embodied ai domains
Embodied & Physical AI
Building intelligent agents that can interact with physical and simulated environments
Reinforcement Learning
Advancing RL algorithms for efficient policy learning and adaptation
Neuro-Symbolic AI
Combining neural networks with symbolic reasoning for robust intelligence
Recent Demos
See our agents navigating complex environments, adapting behaviors, and teaming with humans in the loop.
AI Star Fellowship Support Program
Research and development of cooperative intelligent agents for Human×AI companion collaboration and cultivation of top-tier researchers to lead global excellence
Accelerated Insight Reasoning via Continual Learning
Developing advanced techniques for policy adaptation in reinforcement learning using cross-domain skill diffusion methods for improved generalization
Self-directed Multi-modal Intelligence for Solving Unknown, Open Domain Problems
Securing autonomous learning and adaptation technology for open-domain complex problem situations and demonstrating platform-based agent services Advancing self-supervised learning, meta few-shot learning, and lifelong reinforcement learning for autonomous AI and developing integrated models
Visit CSI-Agent Group
We welcome collaborators and students to drop by our space inside the Suwon Industry-Academia Cooperation Center. Reach out ahead of time to schedule a tour of the lab and robotics arena.
Weekdays · 09:00 – 18:00 (KST)