4 papers
X-Edit: Exact, Explicit, and Explainable Null-Space Editing for Medical Vision Transformers
Yuanye Liu, Siyuan Zhou, Ke Zhang +3
Pre-trained Vision Transformers (ViTs) are increasingly deployed for medical image classification. However, correcting their inevitable failure cases in dynamic clinical scenarios…
Correcting Stochastic Update Bias in Preconditioned Language Model Optimizers
Nikhil Nayak, Julia White, Urchade Zaratiana +7
Preconditioned optimizers are central to language model training, but their stochastic update rules are usually treated as direct approximations to population preconditioned descen…
Multi-Scale Generative Modeling with Heat Dissipation Flow Matching
Jun Ma, Hanquan Zhang, Yanjun Qin +2
Diffusion models are widely used in image generation, with most relying on noise-based corruption and denoising. A distinct branch instead uses blur as the main corruption, preserv…
Semantically Consistent Video Inpainting with Conditional Diffusion Models
Dylan Green, William Harvey, Saeid Naderiparizi +10
Current state-of-the-art methods for video inpainting typically rely on optical flow or attention-based approaches to inpaint masked regions by propagating visual information acros…