IC-888SLIMG fails on heterophily graphs for link prediction because it cannot properly measure node similarity of heterophily embeddings

Meng-Chieh Lee, Haiyang Yu, Jian Zhang, Vassilis N. Ioannidis, Xiang song, Soji Adeshina, Da Zheng, Christos Faloutsos

SourceNetInfoF Framework: Measuring and Exploiting Network Usable Information

The paper reports that SLIMG, a linear GNN designed for node classification, degrades substantially when applied to link prediction on heterophily graphs. On five real-world heterophily datasets (chameleon, squirrel, actor, twitch, pokec), SLIMG scores range from 12.0 to 76.9 hits@100/1000, while on homophily graphs it reaches up to 86.8. On synthetic off-diagonal (heterophily) structures with local features, SLIMG drops from 82.5 to 31.1 hits@100. The authors attribute this to SLIMG's inability to adjust node similarity for embeddings where connected nodes are expected to be dissimilar.

Evidence

correlational

Key metric

SLIMG hits@100: chameleon 76.9±2.8, squirrel 19.6±1.5, actor 18.7±1.0, twitch 12.0±0.3, pokec 21.7±0.2 (heterophily); synthetic local x off-diag a: 31.1±1.1 vs local x diag a: 82.5±1.6

Caveat

The paper notes that SLIMG was designed for node classification, not link prediction, so the failure may be a task mismatch rather than a fundamental architectural limitation.

Model

SLIMG

Concepts

Failure mode

Datasets

Chameleon [eval], Squirrel [eval], Pokec [eval]

Related work

SLIMG [context]

Extraction

automatic-extraction