Qiang Yang

Federated learning struggles when data quality varies across clients, but FedQual solves this with a novel approach that calibrates low-quality clients while preserving high-quality autonomy.

Federated learning can overcome data silos, but struggles when clients have different label relationships; FedHarmony shows how to harmonize these differences, leading to better performance.

Achieve centralized-level performance in federated LLM fine-tuning without compromising IP, privacy, or performance on heterogeneous data by using a compressed "proxy" model.

Seemingly harmless instructions can be weaponized to cause real-world harm by exploiting the limited causal reasoning of embodied LLMs at the action level.