<ul><li>Federated In-Context Learning (Fed-ICL) is proposed as a framework to improve answer quality in question-answering tasks without transmitting model parameters.</li><li>Fed-ICL enhances In-Context Learning (ICL) through iterative refinement via multi-round interactions between clients and a central server.</li><li>It addresses challenges of scarce high-quality examples by leveraging examples stored on client devices.</li><li>Extensive experiments on standard QA benchmarks show that Fed-ICL achieves strong performance with low communication costs.</li></ul>

Federated In-Context Learning: Iterative Refinement for Improved Answer Quality

Discover more