BESplit: Bias-Compensated Split Federated Learning with Evidential Aggregation

2026-05-17

AI Summary

This work addresses the optimization bias and convergence instability in partitioned federated learning under non-IID data, which arise from heterogeneous client representations. To mitigate these issues, the authors propose a novel representation-level bias compensation framework that integrates Evidence Aggregation (EA), Bias Compensation Collaboration (BCC), and Dual-Teacher Distillation (DTD). This approach enables fine-grained reweighting of client contributions and jointly optimizes client and server models to achieve cross-device representation alignment and knowledge synchronization. Extensive experiments on five benchmark datasets demonstrate that the proposed method significantly outperforms existing approaches in terms of accuracy, convergence stability, and computational efficiency.

Abstract

Split Federated Learning (SFL) enables privacy-preserving collaborative training by partitioning models between clients and a server. However, under non-IID data distributions, SFL often suffers from biased optimization and unstable convergence, while existing solutions largely adapt techniques from conventional federated learning. In this work, we observe that the split architecture of SFL inherently alters how client information is represented and coordinated, opening opportunities for bias compensation beyond parameter-level aggregation. Based on this insight, we propose BESplit, an architecture-aware framework that exploits the intrinsic structure of SFL to mitigate non-IID effects. First, to prevent biased local data from dominating global updates, we introduce Evidential Aggregation (EA) to perform fine-grained reweighting of client contributions based on evidential uncertainty. Second, to further reduce distributional skew, we develop Bias-Compensated Collaboration (BCC) to align split-layer representations by pairing complementary clients. Finally, Dual-Teacher Distillation (DTD) is incorporated to synchronize knowledge between decoupled client and server models, enabling independent local inference. Extensive experiments on five benchmark datasets demonstrate that BESplit consistently outperforms state-of-the-art methods in accuracy, convergence stability, and computational efficiency under diverse non-IID settings.

Problem

• Split Federated Learning
• non-IID
• bias
• convergence instability
• privacy-preserving

Innovation

• Split Federated Learning
• Evidential Aggregation
• Bias Compensation
• Non-IID
• Knowledge Distillation