News

SpaCeFormer: Fast Proposal-Free Open-Vocabulary 3D Instance SegmentationChris Choy, Junha Lee, Chunghyun Park, Minsu Cho, Jan KautzInternational Conference on Machine Learning (ICML), Seoul, South Korea, 2026Vision-aligned Latent Reasoning for Multi-Modal Large Language ModelByungwoo Jeon, Yoonwoo J..

Revisiting data imbalance in token-based self-supervised learningDaeyoung Han, Hyung Rok Jung, Tianhong Li, Dina Katabi, Jeany Son, Hong kook Kim, Moongu JeonNeurocomputing, 2026

Prof. Minsu Cho was awarded the Presidential Commendation at the 2026 Korea Science Day ceremony in recognition of outstanding contributions to Artificial Intelligence research.

Hyunmin Kim won the POSTECHIAN fellowship award. Congratulations!

Planning in 8 Tokens: A Compact Discrete Tokenizer for Latent World ModelDongwon Kim, Gawon Seo, Jinsung Lee, Minsu Cho, Suha KwakIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Denver, 2026FlashDecoder: Real-Time Latent-to-Pixel Streaming Decoder with TransformersMinguk Kang,..

Exploring State-Space Models for Data-Specific Neural RepresentationJinsung Lee and Suha Kwak International Conference on Learning Representations (ICLR), Rio de Janeiro, Brazil, 2026

We won six paper awards including the gold prize at IPIU 2026.Gold Prize Event-Centric Chain-of-Thought for Reasoning Temporal Localization in Videos (Youngkil Song, Yoonjae Baek, Dongwon Kim, Inho Kim, Dongkeun Kim, Suha Kwak)Bronze Prize생성적 복원을 통한 RGBD 영상에서의 3차원 행동유도성 인식 학습 (Chunghyun Park, S..

Improving Target Presence and Plurality Recognition for Generalized Referring Image SegmentationNamyup Kim*, Jinsung Lee*, and Suha Kwak (*equal contribution)AAAI Conference on Artificial Intelligence (AAAI), 2026

GaRA-SAM: Robustifying Segment Anything Model with Gated-Rank AdaptationSohyun Lee, Yeho Gwon, Lukas Hoyer and Suha KwakAdvances in Neural Information Processing Systems (NeurIPS), San Diego, 2025Part-Aware Bottom-Up Group Reasoning for Fine-Grained Social Interaction DetectionDongkeun Kim, Minsu Ch..

Enhancing Cost Efficiency in Active Learning with Candidate Set QueryYeho Gwon, Sehyun Hwang, Hoyoung Kim, Jungseul Ok, and Suha KwakTransactions on Machine Learning Research (TMLR), 2025

Profs. Cho and Kwak (Datasets and Benchmarks track) serve as Area Chair (AC) for NeurIPS 2025.

Profs. Cho and Kwak serve as Senior Area Chair (SAC) for ICCV 2025.