Simon Razniewski

Language models may score high on benchmarks, but LLMPEDIA reveals their true factual accuracy is only 68.4%, exposing critical gaps in their encyclopedic knowledge.

GPTKB 2.0 achieves the unprecedented feat of creating a million-scale knowledge base directly from LLMs, complete with disambiguated entities and relations.

LLMs often fail to surface relevant knowledge when prompted openly, revealing a significant gap between what they know and what they readily express.