Guilin Deng

Beyond Client Clustering: Fine-Grained Preference Alignment in Federated RLHF via Self-Evolving Routing

FedPrism is proposed, a novel framework that shifts alignment granularity from the coarse client level to the precise instance level and devise a Posterior-Guided Performance Distillation mechanism that leverages experts' actual training loss as self-supervision to autonomously refine routing policies without explicit...