DANCE: Dynamic, Available, Neighbor-gated Condensation for Federated Text-Attributed Graphs
Zekai Chen, Haodong Lu, Xunkai Li, Henan Sun, Jia Li, Hongchao Qin, Rong-Hua Li, Guoren Wang
TL;DR
This work tackles TAG-FGL when textual attributes are long and costly to process, introducing DANCE, a round-wise condensation framework that is model-in-the-loop and evidence-driven. DANCE combines label-aware node condensation, hierarchical text condensation, and self-expressive topology reconstruction to produce a compact, interpretable propagation graph while preserving privacy via secure aggregation. Theoretical guarantees on budgeted processing and stability accompany empirical results across eight TAG datasets, achieving state-of-the-art accuracy (2.33% gain at 8% condensation) with substantial token reductions (33.42%) per condensed node. The approach enables efficient, auditable federated learning on text-rich graphs with practical privacy protections and interpretable, on-device evidence traces that support regulatory compliance.
Abstract
Federated graph learning (FGL) enables collaborative training on graph data across multiple clients. With the rise of large language models (LLMs), textual attributes in FGL graphs are gaining attention. Text-attributed graph federated learning (TAG-FGL) improves FGL by explicitly leveraging LLMs to process and integrate these textual features. However, current TAG-FGL methods face three main challenges: \textbf{(1) Overhead.} LLMs for processing long texts incur high token and computation costs. To make TAG-FGL practical, we introduce graph condensation (GC) to reduce computation load, but this choice also brings new issues. \textbf{(2) Suboptimal.} To reduce LLM overhead, we introduce GC into TAG-FGL by compressing multi-hop texts/neighborhoods into a condensed core with fixed LLM surrogates. However, this one-shot condensation is often not client-adaptive, leading to suboptimal performance. \textbf{(3) Interpretability.} LLM-based condensation further introduces a black-box bottleneck: summaries lack faithful attribution and clear grounding to specific source spans, making local inspection and auditing difficult. To address the above issues, we propose \textbf{DANCE}, a new TAG-FGL paradigm with GC. To improve \textbf{suboptimal} performance, DANCE performs round-wise, model-in-the-loop condensation refresh using the latest global model. To enhance \textbf{interpretability}, DANCE preserves provenance by storing locally inspectable evidence packs that trace predictions to selected neighbors and source text spans. Across 8 TAG datasets, DANCE improves accuracy by \textbf{2.33\%} at an \textbf{8\%} condensation ratio, with \textbf{33.42\%} fewer tokens than baselines.