English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large language models (LLMs) outperform information retrieval techniques fordownstream knowledge-intensive tasks when being prompted to generate worldknowledge. However, community concerns abound regarding the factuality andpotential implications of using this uncensored knowledge. In light of this, weintroduce CONNER, a COmpreheNsive kNowledge Evaluation fRamework, designed tosystematically and automatically evaluate generated knowledge from siximportant perspectives -- Factuality, Relevance, Coherence, Informativeness,Helpfulness and Validity. We conduct an extensive empirical analysis of thegenerated knowledge from three different types of LLMs on two widely studiedknowledge-intensive tasks, i.e., open-domain question answering andknowledge-grounded dialogue. Surprisingly, our study reveals that thefactuality of generated knowledge, even if lower, does not significantly hinderdownstream tasks. Instead, the relevance and coherence of the outputs are moreimportant than small factual mistakes. Further, we show how to use CONNER toimprove knowledge-intensive tasks by designing two strategies: PromptEngineering and Knowledge Selection. Our evaluation code and LLM-generatedknowledge with human annotations will be released to facilitate futureresearch.

Beyond Factuality: A Comprehensive Evaluation of Large Language Models as Knowledge Generators