Get Started
Research questionHow can LLM-derived knowledge bases disambiguate homonymous entities and preserve auditable fact provenance?LLM-extracted facts may conflate entities that share a name, while opaque extraction makes mistakes difficult to investigate. Querying and reuse become less reliable when users cannot determine which entity a fact refers to or how that fact was produced.
AI
Information Retrieval
Natural Language Processing
Latest papersRecent research connected to this question, newest first.GPTKB 2.0: Browsing, Querying, and Auditing a Disambiguated LLM-Derived Knowledge BaseGPTKB 2.0 is a web-accessible knowledge base containing 38.4 million triples over 1.6 million canonical entities, with consolidated relations and classes. It supports entity browsing, SPARQL queries, natural-language questions translated to SPARQL, and entity linking from user text. Its audit view exposes surface forms, candidate matches, source triples, and disambiguation decisions; the full knowledge base is also available for offline use. The source describes context-guided disambiguation and provenance inspection but provides no comparative accuracy evidence.research paper · Sep 2, 2026
Related questions
How can LLM-built knowledge bases prevent duplicate or conflated entities, relations, and classes at scale?How can we construct accurate knowledge bases from an LLM’s parametric knowledge without retrieval, fine-tuning, or exceeding 32B parameters?How can we evaluate LLM knowledge updates over time without contamination or inconsistent facts?How can we certify LLM-generated, sample-level hypotheses without circular verification or spurious correlations?
Home
Topics
Search
Library