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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Although demonstrating superb performance on various NLP tasks, largelanguage models (LLMs) still suffer from the hallucination problem, whichthreatens the reliability of LLMs. To measure the level of hallucination ofLLMs, previous works first categorize the hallucination according to thephenomenon similarity, then quantify the proportion that model outputs containhallucinatory contents. However, such hallucination rates could easily bedistorted by confounders. Moreover, such hallucination rates could not reflectthe reasons for the hallucination, as similar hallucinatory phenomena mayoriginate from different sources. To address these issues, we propose tocombine the hallucination level quantification and hallucination reasoninvestigation through an association analysis, which builds the relationshipbetween the hallucination rate of LLMs with a set of risk factors. In this way,we are able to observe the hallucination level under each value of each riskfactor, examining the contribution and statistical significance of each riskfactor, meanwhile excluding the confounding effect of other factors.Additionally, by recognizing the risk factors according to a taxonomy of modelcapability, we reveal a set of potential deficiencies in commonsensememorization, relational reasoning, and instruction following, which mayfurther provide guidance for the pretraining and supervised fine-tuning processof LLMs to mitigate the hallucination.

Quantifying and Attributing the Hallucination of Large Language Models via Association Analysis