We build AI that reasons over structure: graphs, knowledge, and multimodal data.
LearnData Lab is led by Prof. Hogun Park in the College of Computing and Informatics at Sungkyunkwan University (SKKU). We develop machine learning models for complex relational data, such as interaction networks, knowledge graphs, and multi-sensor data, and make modern AI systems, from GNNs to LLMs, transparent, robust, and verifiable. Our work appears at leading venues including ICLR, ICML, KDD, WWW, CVPR, and ACL, and is applied to recommendation, healthcare, and industrial AI in collaboration with partners such as Samsung and LG. Prof. Park is currently a Visiting Scholar at UC San Diego’s HDSI (2026-27).
We develop self-supervised, subgraph, and federated approaches to learning from relational structure, with an emphasis on scalability and robustness.
We study complex query answering and multi-hop/numerical reasoning over knowledge graphs, and build structure-grounded LLM systems including GraphRAG and multimodal KG-enhanced generation. Emerging directions include agentic reasoning and orchestration over memories.
We make deep learning models transparent, robust, and verifiable across architectures. For GNNs, we develop post-hoc and counterfactual explanations, influence-based analysis, and adversarial purification; for LLMs and generative models, we study explainability, watermarking, and knowledge editing.
We are looking for passionate new Ph.D. students (or M.S./Ph.D. integrated program students) and Postdocs to join the team. (more info) ! No opening for MS students for now.
We are grateful for funding from the National Research Foundation of Korea (NRF), BK21, Information & Communications Technology Planning & Evaluation (IITP), National IT Industry Promotion Agency (NIPA), Korea Creative Content Agency (KOCCA), LG Electronics, Samsung Research, NCSOFT, Samsung Display, and NAVER.
Two papers have been accepted to the KDD 2026. Congratulations, Geonhee Han, Seungryeol Baek, Heesoo Jung, and Hyunju Kang!
Feb. 2026One paper has been accepted to the ESWA Journal. Congratulations, Jaehyun and Heesoo!
Jan. 2026Two papers have been accepted to the ICLR 2026. Congratulations, Wooseok, Hyunju, and Woohyun!
Aug. 2025A paper has been accepted to the CIKM 2025. Congratulations, Minku!
July 2025I gave a talk titled Providing Explanations for Unsupervised Graph Learning Models at the University of British Columbia.
May 2025A paper has been accepted to the ACL 2025. Congratulations, Hongjun, Minji,and Heesoo!
May 2025A paper has been accepted to the KDD 2025. Congratulations, Heesoo, Chanyong, and Geonhee!
May 2025A paper has been accepted to the ICML 2025. Congratulations, Woohyun!
Feb. 2025A paper has been accepted to the PAKDD 2025. Congratulations, Hongjun, Heesoo, Gayeong, and Juann!