works on
federated learning 1model sparsity 1non-iid data 1representation alignment 1vision models 1

From the 1 of 1 linked paper with an AI index.

collaborators

1 paper

cs.LG2026

Mechanistic Evidence for Preserved-but-Misaligned Representations in Non-IID FedAvg

Muhammad Haseeb, Salaar Masood, Muhammad Abdullah Sohail +2

The paper investigates why federated averaging (FedAvg) performs poorly on non‑IID client data, finding that client models retain useful internal representations but these become m…