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) are claimed to be capable of Natural LanguageInference (NLI), necessary for applied tasks like question answering andsummarization. We present a series of behavioral studies on several LLMfamilies (LLaMA, GPT-3.5, and PaLM) which probe their behavior using controlledexperiments. We establish two biases originating from pretraining which predictmuch of their behavior, and show that these are major sources of hallucinationin generative LLMs. First, memorization at the level of sentences: we showthat, regardless of the premise, models falsely label NLI test samples asentailing when the hypothesis is attested in training data, and that entitiesare used as ``indices'' to access the memorized data. Second, statisticalpatterns of usage learned at the level of corpora: we further show a similareffect when the premise predicate is less frequent than that of the hypothesisin the training data, a bias following from previous studies. We demonstratethat LLMs perform significantly worse on NLI test samples which do not conformto these biases than those which do, and we offer these as valuable controlsfor future LLM evaluation.

Sources of Hallucination by Large Language Models on Inference Tasks