Multimodal graph model improves predictions using geometric algebra
ICE: Task-Aligned Clifford Latent Fields for Multimodal Graph Foundation Models
Machine Learning
Summary
Graphs that connect things like text, images, and relationships are tricky to understand all at once. The paper’s authors created a new model, ICE, that uses special math called Clifford algebra to better capture how information flows and interacts in these complex graphs. This approach helps the model keep the meaning of each entity clear while combining different types of data smoothly. Their tests show it works better than other methods on many tasks involving prediction from graphs with text, images, and connections.
What this means in practice
- •For machine learning engineers: Build models that analyze graph data with multiple modalities like images and text for better prediction accuracy.
- •For social network analysts: Improve relationship and content understanding in social networks by capturing richer interactions within multimodal connections.
Authors
Xunkai Li, Xu Wang, Yinlin Zhu, Xiong Yongfu, Yi Liu, Rong-Hua Li, Guoren Wang
Abstract
Multimodal attributed graphs connect entities, visual content, language, and observed relations. Learning one foundation across such graphs requires more than compressing each node into a fused Euclidean vector. The representation must preserve entity semantics, construct interaction state from graph neighborhoods, and expose that state to prediction units with different geometry. Our empirical study shows why these requirements are inseparable. Higher-grade channels recover pair relations across the foundation graphs, specialized queries reveal information hidden by a generic readout, and rigid blade isolation removes cross-grade capacity. We therefore introduce ICE (Interaction-aware Clifford Encoder), a multimodal graph foundation model built on a node-indexed Clifford latent field. Topology, text, and images enter explicit Cl(3) addresses. Edge-aware geometric products transform these directions into scalar, bivector, and trivector relations over observed neighborhoods. A protected Grade-1 route preserves entity semantics, while the full grade and depth bank remains available to fresh node and link heads. We establish exact cross-grade reachability, node-permutation equivariance, and a bound on the task residual around the semantic score. Experiments span one shared foundation over eleven graphs, six node-classification datasets, three link-prediction datasets, and matched few-shot tasks. ICE ranks first in all 30 reported supervised and few-shot comparisons. Core removals reduce every task summary, and mechanism controls connect the gains to higher-order transport, retained multidepth structure, semantic protection, and direct field access.