Language-encoded network topology enables large language models to reason about complex networks
Across twenty networks spanning five domains, BioGlyph substantially improves open LLMs'ability to answer structural reasoning questions, outperforming edge-based, numerical and learned representations by up to 26 percentage points in system accuracy.
Ucchwas Talukder Utsha, Sakib Mostafa, James Zou et al.
· 0 citations