Search papers, labs, and topics across Lattice.
This paper introduces SDM-SCR, a novel graph contrastive learning framework for text-attributed graphs that uses LLMs to disentangle task-relevant signals from noise via approximate orthogonal decomposition. The Semantic Decoupling Module (SDM) leverages LLMs to parse raw attributes into signal and noise views, while Semantic Consistency Regularization (SCR) acts as a spectral filter to enforce consistency on the signal subspace. Experiments show SDM-SCR achieves state-of-the-art performance in accuracy and efficiency by purifying signals and reducing LLM hallucinations.
LLMs can actively disentangle signal from noise in text-attributed graphs, leading to state-of-the-art performance in graph contrastive learning by shifting from random perturbation to semantic-aware disentanglement.
Conventional Graph Contrastive Learning (GCL) on Text-Attributed Graphs (TAGs) relies on blind stochastic augmentations, inadvertently entangling task-relevant signals with noise. We propose SDM-SCR, a robust framework anchored in Approximate Orthogonal Decomposition. First, the Semantic Decoupling Module (SDM) leverages the instruction-following capability of Large Language Models (LLMs) to actively parse raw attributes into asymmetric, task-oriented signal and noise views. This shifts the paradigm from random perturbation to semantic-aware disentanglement. Subsequently, Semantic Consistency Regularization (SCR) exploits the spectral observation that semantic signals are topologically smooth while residual noise is high-frequency. SCR functions as a selective spectral filter, enforcing consistency only on the signal subspace to eliminate LLM hallucinations without over-smoothing. This ``Disentangle-then-Refine''mechanism ensures rigorous signal purification. Extensive experiments demonstrate that SDM-SCR achieves SOTA performance in accuracy and efficiency.
