SLM-Conditioned Routing Integrates Small Language Models into Property Graph Learning
August 28, 2026
This architecture uses a parameter-efficient small language model to generate target-conditioned routing queries within a GNN. The SLM processes structured graph tokens to dynamically select relevant semantic messages for node prediction in labeled property graphs.
HOW THIS AFFECTS YOU
●
researcherYou can improve graph neural networks by using SLMs to handle heterogeneous node and relationship properties dynamically.