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

By providing external information to large language models (LLMs), toolaugmentation (including retrieval augmentation) has emerged as a promisingsolution for addressing the limitations of LLMs' static parametric memory.However, how receptive are LLMs to such external evidence, especially when theevidence conflicts with their parametric memory? We present the firstcomprehensive and controlled investigation into the behavior of LLMs whenencountering knowledge conflicts. We propose a systematic framework to elicithigh-quality parametric memory from LLMs and construct the correspondingcounter-memory, which enables us to conduct a series of controlled experiments.Our investigation reveals seemingly contradicting behaviors of LLMs. On the onehand, different from prior wisdom, we find that LLMs can be highly receptive toexternal evidence even when that conflicts with their parametric memory, giventhat the external evidence is coherent and convincing. On the other hand, LLMsalso demonstrate a strong confirmation bias when the external evidence containssome information that is consistent with their parametric memory, despite beingpresented with conflicting evidence at the same time. These results poseimportant implications that are worth careful consideration for the furtherdevelopment and deployment of tool- and retrieval-augmented LLMs.

Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts