Are Large Language Models Suitable for Graph Computation? Progress and Prospects

In three linesComprehensive review of LLMs for graph computation. Authors propose role-based taxonomy: LLMs as executors (direct task solving) and LLMs as planners (problem formulation, reasoning decomposition, external tool invocation). Finding: LLMs reliable for simple, small-scale tasks but unreliable for large-scale and exactness-demanding tasks.

Summary generated by Claude — human-verified