Reproduce “Fine-Tuning Impairs the Balancedness of Foundation Models in Long-tailed Personalized Federated Learning” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Shihao Hou, Chikai Shang, Zhiheng Yang, Jiacheng Yang, Xinyi Shang, Junlong Gao, Yiqun Zhang, Yang Lu

FedPuReL protects a shared model’s class balance during federated training and personalizes it with client-specific additive logit corrections. It uses zero-shot predictions to remove conflicting shared-model updates, then freezes that model while clients train local branches. On CIFAR-100-LT at imbalance factor 500, it reports 72.91% global and 77.73% personalized accuracy, versus 70.32% and 74.17% for CLIPLoRA+Fed-GraB.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.