GPTKB v2

A large general-domain knowledge base entirely from a large language model.

Browse all Dataset profile SPARQL query Download

38.5 million triples 1.3 million entities

96% entity precision 100% homonymy 91% synonymy resolved

About

GPTKB v2 is a general-domain knowledge base built from GPT-5 — 38.5 million triples across 1.3 million disambiguated entities. The browser exposes the disambiguation chain (surface form → canonical entity → provenance batch) on every triple.

Try an entity

Example SPARQL queries

Main papers

Two papers on GPTKB v2 are in preparation and will appear on arXiv shortly. The Publications page lists the earlier GPTKB work this builds on.

In preparation
GPTKB v2 — the knowledge base
To appear on arXiv
In preparation
GPTKB v2 — the web interface
To appear on arXiv