Or Patashnik

DeLeaker: Dynamic Inference-Time Reweighting For Semantic Leakage Mitigation in Text-to-Image Models

VLM-Guided Adaptive Negative Prompting for Creative Generation

Kontinuous Kontext: Continuous Strength Control for Instruction-based Image Editing

Scaling Group Inference for Diverse and High-Quality Generation

SAEdit: Token-level control for continuous image editing via Sparse AutoEncoder

NoisePrints: Distortion-Free Watermarks for Authorship in Private Diffusion Models

Visual Diffusion Models are Geometric Solvers

Image Generation from Contextually-Contradictory Prompts

Noise-free Score Distillation