AlignFed Tackles Straggler Effect in Async Federated LLM Fine-Tuning

AlignFed introduces alignment-aware asynchronous federated fine-tuning targeting heterogeneous edge deployments where synchronous aggregation causes straggler bottlenecks. The method addresses non-IID data and resource heterogeneity specific to LLM-scale models, unlike prior async FL work designed for smaller models.