Get Started
Research questionHow can we construct accurate knowledge bases from an LLM’s parametric knowledge without retrieval, fine-tuning, or exceeding 32B parameters?Language models may encode useful facts without exposing them as complete, correctly typed relations. Extracting those facts requires handling both supported and absent relations while producing valid structured records.
AI
Evaluation & Benchmarks
Natural Language Processing
Latest papersRecent research connected to this question, newest first.REAP: Relation-Aware Elicitation and Parsing for Closed-Book Knowledge Base Construction from LLMsThe source addresses the AKBC Shared Task 2026 setting, which prohibits external retrieval and fine-tuning and limits models to 32B parameters. Reported evidence is a test-set result for a system built on Mistral-Small-24B-Instruct-2501: macro-F1 0.62, with relation-level F1 values of 0.95 for countryLandBordersCountry, 0.73 for companyTradesAtStockExchange, and 0.77 for hasArea.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 LLM-derived knowledge bases disambiguate homonymous entities and preserve auditable fact provenance?How can instruction-tuned LLMs learn corpus-specific knowledge without exhaustive synthetic QA or instruction fine-tuning?How can a model select the correct entity-specific parametric memory for graph queries without query-time text or semantic similarity?
Home
Topics
Search
Library