🇰🇷 🇺🇸 🇦🇺 🇸🇬 🇯🇵 🇲🇳 🇨🇳 🇲🇾 🇶🇦 🇷🇼 🇰🇪 🇻🇳 🇫🇷 🇮🇹 🇩🇪 🇬🇧 🇳🇱 🇵🇹 🇪🇸

• Model Already Knows the Best Noise: Bayesian Active Noise Selection via Attention in Video Diffusion Model [paper]
  Kwanyoung Kim, Sanghyun Kim
  arXiv preprint.
• Safety Alignment Backfires: Preventing the Re-emergence of Suppressed Concepts in Fine-tuned Text-to-Image Diffusion Models [paper]
  Sanghyun Kim, Moonseok Choi, Jinwoo Shin, Juho Lee
  arXiv preprint.
• Safeguard Text-to-Image Diffusion Models with Human Feedback Inversion [paper] [code]
  Sanghyun Kim, Seohyeon Jung, Balhae Kim, Moonseok Choi, Jinwoo Shin, Juho Lee
  ECCV 2024 (Acceptance rate: 27.9%)
• Slot-Mixup with Subsampling: A Simple Regularization for WSI Classification [arXiv]
  Seongho Keum, Sanghyun Kim, Soojeong Lee, Juho Lee
  ariXiv preprint.
• Towards safe self-distillation of Internet-scale text-to-image diffusion models [paper] [code]
  Sanghyun Kim, Seohyeon Jung, Balhae Kim, Moonseok Choi, Jinwoo Shin, Juho Lee
  ICML 2023 Workshop on Challenges of Deploying Generative AI
• Modeling uplift from observational time-series in continual scenarios [paper] [code] [PMLR]
  Sanghyun Kim, Jungwon Choi, NamHee Kim, Jaesung Ryu, Juho Lee
  AAAI 2023 Bridge on Continual Casality (oral presentation), 2023
• A simple yet powerful deep active learning with snapshot ensembles [paper] [code]
  Seohyeon Jung*, Sanghyun Kim*, Juho Lee
  (*: Equal Contribution)
  ICLR 2023 (Acceptance rate: 24.3%)