
CleaR: Robust and Generalized Parameter-Efficient Fine-Tuning for Noisy Labels
PaulHoule
arxiv.org
23 points0 comments
Summary
by metafa.stThis paper presents a novel approach to generating high-fidelity images from text prompts using a diffusion model trained on a large dataset of aligned images and captions. The method demonstrates competitive performance compared to existing models while achieving improved efficiency and reduced computational requirements.