BESplit: Bias-Compensated Split Federated Learning with Evidential Aggregation

TL;DR

BESplit is a novel framework for split federated learning that effectively mitigates non-IID data biases through architecture-aware strategies, evidential aggregation, and dual-teacher distillation, improving accuracy and stability.

Contribution

The paper introduces BESplit, a new architecture-aware approach that leverages the intrinsic structure of SFL to address bias and convergence issues under non-IID data distributions.

Findings

BESplit outperforms existing methods in accuracy across five benchmark datasets.

It achieves more stable convergence and better computational efficiency.

The framework effectively reduces bias caused by non-IID data in split federated learning.

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.