1 paper · 1 filter
Beining Wu, Zihao Ding, Jun Huang
While current federated multimodal continual learning over mixture-of-experts low-rank adaptation (MoE-LoRA) is built on the unverified assumption that routing isolates task-specif…