LLMs have majorly advanced NLP and AI, and next to their ability to perform a wide range of procedural tasks, a major success factor is their internalized factual knowledge. Since the studies of Petroni et al., analyzing this knowledge has gained attention. However, most approaches investigate one question at a time via modest-sized pre-defined samples, introducing an availability bias that prevents the discovery of knowledge (or beliefs) of LLMs beyond the experimenter's predisposition. To address this challenge, we propose to comprehensively materialize an LLM's factual knowledge through recursive querying and result consolidation. As a prototype, we employ GPT-4o-mini to construct GPTKB, a large-scale knowledge base (KB) comprising 101 million triples for over 2.9 million entities - achieved at a fraction of the cost of previous KB projects. This contributes to two discourses: for LLM research, for the first time, it provides constructive insights into the scope and structure of LLMs' knowledge (or beliefs); for KB construction, it pioneers new pathways for the long-standing challenge of general-domain KB construction.

Simon Razniewski is professor for knowledge-aware AI at ScaDS.AI and TU Dresden, where he develops novel methods for extracting and consolidating commonsense and world knowledge from and with language models and knowledge bases. He was previously a research scientist at the Bosch Center for AI, senior researcher at the Max Planck Institute for Informatics, and assistant professor at the Free University of Bozen Bolzano. He co-organizes the KBC-LM and Wikidata workshop series at ISWC, and regularly publishes at conferences in the fields of Computational Linguistics (ACL, EMNLP), Semantic Web (ISWC, WWW) and Knowledge Management (WSDM, CIKM).