Abstract page for arXiv paper 2611.18638: GRAFT: Cross-Backbone LoRA Reuse for Continual Language and Multimodal Learning