LLM-generated node embeddings improve GCN and Graph Transformer node classification on Cora and PubMed in high-label splits, but gains shrink or reverse under low-label splits.
A comprehensive study on tex t-attributed graphs: Benchmarking and rethinking,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Deep Semantic Graph Learning via LLM based Node Enhancement
LLM-generated node embeddings improve GCN and Graph Transformer node classification on Cora and PubMed in high-label splits, but gains shrink or reverse under low-label splits.