We encourage submissions exploring theoretical insights into vision foundation models (VFMs) and large language models, efficient foundation model architectures, and hybrid network designs combining VFMs with convolutional and graph-based models. Additionally, contributions addressing challenges and strategies involved in leveraging VFMs and LMMs for tasks such as image and video generative models, unsupervised, weakly, and semi-supervised learning settings, multi-modal foundation models, and improving semantic understanding of visual content in multi-spectral data are welcomed. The workshop also aims to accelerate the adoption of foundation models in various industrial applications by investigating inherent biases, blind spots, and strategies for safe and privacy-preserving deployment. We welcome papers reporting experimental results on accelerating the training and inference time of foundation models, as well as their adaptation for low-level and high-level vision problems, mobile devices, and embodied AI.