AI RESEARCH
CRAM: Centroid-Routing and Adaptive MoE for Multimodal Continual Instruction Tuning
arXiv CS.CL
•
ArXi:2606.02502v1 Announce Type: new Multimodal Large Language Models (MLLMs) unify heterogeneous vision-language tasks under a shared generative framework via instruction tuning, yet real-world deployment demands continuous capability expansion, making Multimodal Continual Instruction Tuning (MCIT) essential. Existing methods either update all tasks with a shared parameter set or allocate dedicated modules for each new task. Shared updates force heterogeneous tasks to compete, causing forgetting of learned capabilities.