OLIVE: Object Level In-Context Visual Embeddings

Timothy Ossowski, Junjie Hu

Recent generalist vision-language models (VLMs) have demonstrated impressive reasoning capabilities across diverse multimodal tasks. However, these models still struggle with fine-…