When Is Shallow Enough? Adaptive Split Federated Learning with Client-Specific Sufficiency Estimation

2026-08-16

AI Summary

This study addresses the suboptimality of static splitting in split federated learning caused by client heterogeneity by proposing the FedSGA framework. The method leverages private prompt tracking for dynamic adaptation, combines semantic alignment with stability estimation to evaluate shallow layer sufficiency, and employs an interface coordination module to unify heterogeneous feature spaces, thereby achieving adaptive splitting. Experimental results demonstrate that FedSGA outperforms existing methods across multiple heterogeneous benchmarks. It effectively enhances model accuracy while significantly reducing computational redundancy on clients, providing an efficient solution for adaptive splitting in heterogeneous environments.

Abstract

\textit{Split Federated Learning} (SFL) enables distributed model training by splitting networks between the server and clients. However, under client heterogeneity, the conventional static split strategy may be suboptimal because clients can differ in data distributions, adaptation dynamics, and representation learning progress, making a single split point insufficient to accommodate client-specific training states. In this paper, we propose \textsc{FedSGA}, a \textbf{S}ufficiency-\textbf{G}uided \textbf{A}daptive split \textbf{Fed}erated learning framework that addresses this question through client-specific shallow sufficiency estimation. First, we introduce a client-specific adaptation channel based on private prompt tokens, which tracks local adaptation dynamics separately from the shared backbone and provides a lightweight signal for detecting whether client adaptation remains active. To further avoid repeated online probing over multiple candidate depths, we design a shallow sufficiency estimator that combines cross-client semantic alignment, temporal interface stability, and prompt-state variation to estimate whether the shallowest split is already sufficient. Finally, we introduce a split-compatible interface harmonization module that projects activations from different split depths into a shared semantic space, improving the comparability of heterogeneous client interfaces before server-side prediction. Extensive experiments on multiple heterogeneous benchmarks demonstrate the effectiveness of \textsc{FedSGA} in improving model performance compared with state-of-the-art methods while reducing unnecessary client-side computation.

Problem

• Split Federated Learning
• Client Heterogeneity
• Static Split Strategy
• Adaptive Splitting

Innovation

• Adaptive Split Federated Learning
• Shallow Sufficiency Estimation
• Private Prompt Tokens
• Interface Harmonization
• Client Heterogeneity