Publication
Latent Inversion with Timestep-aware Sampling for Training-free Non-rigid Editing
Feb 13, 2024 · arXiv (Cornell University)
Tair Djanibekov
@tair-djanibekov
PhD Student · MBZUAI
I'm interested in semantic watermarking, which makes AI-generated text traceable by its meaning so the signal survives paraphrasing, and in scalable oversight, which asks how weaker supervisors can reliably evaluate and steer models stronger than themselves.
Open to
Research collaborationStartup collaborationJoining pet projectsPaper discussions