← back to paper
arxiv: 2605.13475 · 2 revisions
FedHPro: Federated Hyper-Prototype Learning via Gradient Matching