Brief

Researchers have introduced SPPE, a new benchmark for evaluating privacy-preserving image editing in Multimodal Large Language Models (MLLMs). This benchmark addresses the issue where standard privacy methods often result in edited surrogate images rather than the desired edited source images. SPPE includes tasks for assessing editability before cloud interaction and for recovering the edited source image from the surrogate, along with novel methods ERMA and C2E-S2SER to tackle these challenges. AI

Apple's research paper explores the mechanisms behind compositional generalization in conditional diffusion models, particularly focusing on how these models handle generating images with more objects than trained on. The study identifies 'local conditional scores' as a key factor enabling this ability, demonstrating that models succeeding at length generalization exhibit these scores, while those that fail do not. The research also proposes a method to enforce these local scores, which successfully enabled length generalization in a previously underperforming model. AI