Dahye Kim

Ph.D. Student in Computer Science ยท Boston University

Dahye Kim

Hi, thanks for stopping by! ๐Ÿ‘‹

I am a third-year CS PhD student at Boston University, where I am fortunate to be advised by Prof. Deepti Ghadiyaram. Prior to joining BU, I completed my Masterโ€™s degree at Yonsei University under the guidance of Prof. Kwanghoon Sohn.

Iโ€™m interested in understanding how generative models work and using those insights to make them more useful, efficient, and controllable.

Iโ€™m always open to collaborations and research discussions. Feel free to reach out!

News

Swift Sampling is accepted at NeurIPS 2026!
Excited to join Adobe as a Research Intern in Seattle! ๐ŸŽจ
DDiT is selected as a Highlight at CVPR 2026! ๐ŸŒŸ
DDiT is accepted at CVPR 2026! ๐Ÿž๏ธ
I joined Amazon as a Research Intern in Seattle! ๐Ÿ™๏ธ
Revelio is accepted at ICCV 2025! ๐ŸŒบ
Revelio is selected for an oral presentation at the CVPR 2025 MIV Workshop ! ๐ŸŽธ
New preprint is out: check out Concept Steerers ๐Ÿช„
New preprint is out: check out Revelio ๐Ÿ”Ž
๐Ÿš€ I started my PhD journey at Boston University!

Publications & Preprints

  1. Swift Sampling: Selecting Temporal Surprises via Taylor Series
    Swift Sampling: Selecting Temporal Surprises via Taylor Series
    Dahye Kim, Bhuvan Sachdeva*, Karan Uppal*, Naman Gupta*, Vineeth N. Balasubramanian, and Deepti Ghadiyaram
    NeurIPS, 2026
    Paper Project Code
  2. DDiT: Dynamic Patch Scheduling for Efficient Diffusion Transformers
    DDiT: Dynamic Patch Scheduling for Efficient Diffusion Transformers
    Dahye Kim, Deepti Ghadiyaram, and Raghudeep Gadde
    CVPR, 2026 (Highlight)
    Paper Project
  3. Concept Steerers: Leveraging K-Sparse Autoencoders for Test-Time Controllable Generations
    Concept Steerers: Leveraging K-Sparse Autoencoders for Test-Time Controllable Generations
    Dahye Kim and Deepti Ghadiyaram
    arXiv, 2025
    Paper Project Code
  4. Revelio: Interpreting and leveraging visual semantic information in diffusion models
    Revelio: Interpreting and leveraging visual semantic information in diffusion models
    Dahye Kim*, Xavier Thomas*, and Deepti Ghadiyaram
    ICCV, 2025
    CVPRW MIV, 2025 (Oral)
    Paper Project Code Poster
  5. Language-free training for zero-shot video grounding
    Language-free training for zero-shot video grounding
    Dahye Kim, Jungin Park, Jiyoung Lee, Seongheon Park, and Kwanghoon Sohn
    WACV, 2023 (Oral)
    Paper
  6. Normality guided multiple instance learning for weakly supervised video anomaly detection
    Normality guided multiple instance learning for weakly supervised video anomaly detection
    Seongheon Park, Hanjae Kim, Minsu Kim, Dahye Kim, and Kwanghoon Sohn
    WACV, 2023 (Oral)
    Paper

Research Experience

Awards and Honors

  • Hariri Institute Student Excellence Award 2026
    Boston University

    Awarded for significant contributions within the Hariri Institute and a commitment to excellence.

Professional Service

  • Workshop Organizer CVPR 2025, CVPR 2026
    XAI4CV: Explainable Artificial Intelligence for Computer Vision
  • Reviewer
    Conferences: CVPR (2026), ECCV (2026), NeurIPS (2026), AAAI (2027).
    Journals: IJCV.

Teaching

Miscellaneous

My Korean name is ๊น€๋‹คํ˜œ (้‡‘่Œถๆ…ง). Dahye means โ€œbeautiful and wise.โ€ Itโ€™s pronounced โ€œdah-hehโ€ ().

Languages: Korean (native), English (fluent).