H4G: Unlocking Faithful Inference for Zero-Shot Graph Learning in Hyperbolic Space
This addresses a bottleneck in graph-text alignment for zero-shot learning, offering a novel method to improve accuracy on heterophilic and homophilic graphs, though it is incremental as it builds on existing hyperbolic space approaches.
The paper tackled the problem of over-abstraction in zero-shot graph learning on text-attributed graphs, particularly for fine-grained pattern recognition on heterophilic graphs, by proposing H4G to reduce embedding radii in hyperbolic space, resulting in state-of-the-art performance with 12.8% improvement on heterophilic graphs and 8.4% on homophilic graphs.
Text-attributed graphs are widely used across domains, offering rich opportunities for zero-shot learning via graph-text alignment. However, existing methods struggle with tasks requiring fine-grained pattern recognition, particularly on heterophilic graphs. Through empirical and theoretical analysis, we identify an \textbf{over-abstraction problem}: current approaches operate at excessively large hyperbolic radii, compressing multi-scale structural information into uniform high-level abstractions. This abstraction-induced information loss obscures critical local patterns essential for accurate predictions. By analyzing embeddings in hyperbolic space, we demonstrate that optimal graph learning requires \textbf{faithful preservation} of fine-grained structural details, better retained by representations positioned closer to the origin. To address this, we propose \textbf{H4G}, a framework that systematically reduces embedding radii using learnable block-diagonal scaling matrices and Möbius matrix multiplication. This approach restores access to fine-grained patterns while maintaining global receptive ability with minimal computational overhead. Experiments show H4G achieves state-of-the-art zero-shot performance with \textbf{12.8\%} improvement on heterophilic graphs and \textbf{8.4\%} on homophilic graphs, confirming that radius reduction enables faithful multi-scale representation for advancing zero-shot graph learning.