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

While Large Language Models (LLMs) have proven to be exceptional on a varietyof tasks after alignment, they may still produce responses that contradict thecontext or world knowledge confidently, a phenomenon known as``hallucination''. In this paper, we demonstrate that reducing theinconsistency between the external knowledge encapsulated in the training dataand the intrinsic knowledge inherited in the pretraining corpus could mitigatehallucination in alignment. Specifically, we introduce a novel knowledgeconsistent alignment (KCA) approach, which involves automatically formulatingexaminations based on external knowledge for accessing the comprehension ofLLMs. For data encompassing knowledge inconsistency, KCA implements severalsimple yet efficient strategies for processing. We illustrate the superiorperformance of the proposed KCA approach in mitigating hallucinations acrosssix benchmarks using LLMs of different backbones and scales. Furthermore, weconfirm the correlation between knowledge inconsistency and hallucination,signifying the effectiveness of reducing knowledge inconsistency in alleviatinghallucinations. Our code, model weights, and data are public at\url{https://github.com/fanqiwan/KCA}.

Mitigating Hallucinations of Large Language Models via Knowledge Consistent Alignment