TL;DR

InstructGLM serializes graph topology and node features as natural language instructions, unifying node classification, link prediction, and graph classification under a single LLM and matching or exceeding GNN baselines in zero-shot and few-shot settings.

Summary

Graph machine learning tasks are typically solved with graph neural networks requiring task-specific architectures. InstructGLM serializes graph topology and node features as natural language instructions, reformulating all graph tasks as text generation problems solvable by a single LLM. On Cora, CiteSeer, and ogbn-arxiv benchmarks, InstructGLM matches or exceeds GNN baselines in zero-shot and few-shot regimes, demonstrating that language models can serve as general-purpose graph reasoners without graph-specific inductive bias.

Key contributions

  1. Proposes InstructGLM, which serializes graph structure and node features as natural language for LLM-based graph task solving.
  2. Unifies node classification, link prediction, and graph classification under a single text-generation framework without task-specific architectures.
  3. Demonstrates zero-shot and few-shot performance that matches or exceeds GNN baselines on standard benchmarks including ogbn-arxiv.
  4. Shows that instruction tuning enables zero-shot transfer across graph task types without retraining.

When to cite

  1. When arguing that LLMs can replace graph-specific architectures when topology is serialized as natural language.
  2. When discussing the use of foundation models to reduce specialized graph ML engineering across downstream tasks.
  3. When citing evidence of zero-shot transfer from instruction-tuned LLMs to graph reasoning tasks.
  4. When exploring the limits of context-length as the primary bottleneck for LLM-based graph processing at scale.

LLM KG Knowledge Graph NLP