Language-Guided Trajectory Traversal in Disentangled Stable Diffusion Latent Space

MIV Workshop @ CVPR 2025

🌐 Project Page  ·  📄 Paper (CVF)  ·  📄 arXiv  ·  🖼️ Poster  · 

TL;DR
This project studies how language-guided fine-tuning of Stable Diffusion leads to disentangled latent directions in medical image generation. We introduce a framework for non-linear latent trajectory traversal that enables controlled, factorized edits (e.g., pathology severity or medical devices) while preserving patient identity. We also propose CFRT (Classifier Flip Rate along a Trajectory) as a quantitative measure of disentanglement.

Citation

@inproceedings{tehraninasab2025language,
  title     = {Language-Guided Trajectory Traversal in Disentangled Stable Diffusion Latent Space for Factorized Medical Image Generation},
  author    = {TehraniNasab, Zahra and Kumar, Amar and Arbel, Tal},
  booktitle = {Proceedings of the Mechanistic Interpretability Workshop (MIV) at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year      = {2025}
}