Active Prompt Learning with Vision-Language Model Priors
Year: 2025 / Author: Hoyoung Kim, Seokhee Jin, Changhwan Sung, Jaechang Kim, Jungseul Ok / Journal: TMLR
Search
Close this search box
Beyond Scaling Laws
Beyond Scaling Laws
The Lab aims for novel model architectures and training algorithms to go beyond the limits of machine learning scaling laws. This research is dedicated to achieving efficient and practical AI by overcoming the high costs and environmental challenges associated with scaling large models.
Publications List
Retrieval-Augmented Generation with Estimation of Source Reliability
Year: 2025 / Author: Jeongyeon Hwang, Junyoung Park, Hyejin Park, Dongwoo Kim, Sangdon Park, Jungseul Ok / Conference: Conference on Empirical Methods in Natural Language Processing (EMNLP)
Adaptive Non-uniform Timestep Sampling for Accelerating Diffusion Model Training
Year: 2025 / Author: Myunsoo Kim, Donghyeon Ki, Seong-Woong Shim, Byung-Jun Lee / Conference: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
K/DA: Automated Data Generation Pipeline for Detoxifying Implicitly Offensive Language in Korean
Year: 2025 / Author: Minkyeong Jeon, Hyemin Jeong, Yerang Kim, Jiyoung Kim, Jae Hyeon Cho, Byung-Jun Lee / Conference: Annual Meeting of the Association for Computational Linguistics (ACL)
Iterative Prompt Refinement for Safer Text-to-Image Generation
Year: 2025 / Author: Jinwoo Jeon, JunHyeok Oh, Hayeong Lee, Byung-Jun Lee / Conference: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)
Rethinking DPO: The Role of Rejected Responses in Preference Misalignment
Year: 2025 / Author: Jay Hyeon Cho, JunHyeok Oh, Myunsoo Kim, Byung-Jun Lee / Conference: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)
The Heart of AI Innovation in Korea
108, Taebong-ro, Seocho-gu, Seoul 06764, Republic of Korea
56 Yangjae-daero 12-gil, Seocho-gu, Seoul 06804, Republic of Korea
Tel. +82 2-577-5722
© National AI Research Lab 2026. All rights reserved.
If you have any problems with the website or need technical support, please contact inquiry@dvn.ci