News

Our lab has been selected as a Basic Research Laboratory (BRL) by the National Research Foundation of Korea (NRF).

Project: Next Generation Hyper-perceptive Hyper-adaptive Action Foundation Model Lab Funding period: July 2026 – June 2029

The Basic Research Laboratory program supports sustained, collaborative basic research, and this grant will fund our work toward next-generation action foundation models that perceive richly across modalities and adapt flexibly to new tasks and environments.

Our collaborative research with the University of Wisconsin-Madison on active speaker detection has been accepted to Interspeech 2026. This is a substantially revised version of our earlier work, UniTalk, retitled to better reflect its focus on in-the-wild generalization and robustness.

Title: Revisiting Active Speaker Detection: An In-the-Wild Benchmark for Generalization and Robustness Authors: Le Thien Phuc Nguyen, Zhuoran Yu, Khoa Quang Nhat Cao, Yuwei Guo, Tu Ho Manh Pham, Tuan Tai Nguyen, Toan Ngo Duc Vo, Lucas Poon, Soochahn Lee, Yong Jae Lee

We look forward to presenting our work at Interspeech 2026!

We have released an initial version of code for our Scientific Reports 2025 paper on interactive segmentation of vascular structures.

Title: Improving retinal vessel assessment precision by integrating deep learning with interactive editing and graphical modeling Authors: Sojung Go, Jaemin Chae, Uichan Kim, Jongsoo Lim, Jooyoung Kim, Sang Jun Park, Soochahn Lee Code: github.com/snubhretina/SERVAL Paper: Scientific Reports article

The released SERVAL (SEoul Retinal Vessel Assessment Library) code provides the GUI tool together with the deep learning vessel segmentation and artery/vein classification modules, the interactive semi-automatic editing operations for vessel connection and A/V label correction, and the graphical modeling step that enforces topological consistency of the vessel tree. We hope this will support the community in building on our work.

Our collaborative research with the University of Wisconsin-Madison has been accepted to ICML 2026.

Title: DocHop: Benchmarking Out-of-domain Multi-hop Reasoning in Information-Dense Documents Authors: Zhuoran Yu, Le Thien Phuc Nguyen, Jaden Park, Xinyi Gu, Zexue He, Soochahn Lee, Rogerio Feris, Yong Jae Lee

We look forward to presenting our work at ICML 2026!

We have released the official code for our MICCAI 2025 paper on interactive segmentation of vascular structures.

Title: Multimodal Prompt Sequence Learning for Interactive Segmentation of Vascular Structures Authors: Jongsoo Lim, Soochahn Lee Code: github.com/kmu-ee-ailab/mpsl-vessel-seg Paper: MICCAI 2025 proceedings

The released code covers training and inference for our prompt sequence model, which combines click- and text-based user interactions with SAM/CLIP encoders to enable accurate interactive segmentation of thin, branching vascular structures. We hope this will support the community in building on our work.

Our collaborative research with the University of Wisconsin-Madison on benchmarking multimodal LLMs has been accepted to CVPR Findings 2026.

Title: Contamination Detection for VLMs Using Multi‑Modal Semantic Perturbations Authors: Le Thien Phuc Nguyen, Zhuoran Yu, Samuel Low Yu Hang, Subin An, Jeongik Lee, Yohan Ban, SeungEun Chung, Thanh-Huy Nguyen, JuWan Maeng, Soochahn Lee, Yong Jae Lee

We look forward to presenting our work at ICLR 2026!

Soochahn Lee will be serving as an Area Chair for MICCAI 2026. In this role, he looks forward to evaluating the latest advancements in medical image computing and contributing to the success of the conference.

Our collaborative research with the University of Wisconsin-Madison has been accepted to ICLR 2026.

Title: Contamination Detection for VLMs Using Multi‑Modal Semantic Perturbations Authors: Jaden Park, Mu Cai, Feng Yao, Jingbo Shang, Soochahn Lee, Yong Jae Lee

We look forward to presenting our work at ICLR 2026!