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) have led to significant improvements in manytasks across various domains, such as code interpretation, response generation,and ambiguity handling. These LLMs, however, when upgrading, primarilyprioritize enhancing user experience while neglecting security, privacy, andsafety implications. Consequently, unintended vulnerabilities or biases can beintroduced. Previous studies have predominantly focused on specific versions ofthe models and disregard the potential emergence of new attack vectorstargeting the updated versions. Through the lens of adversarial examples withinthe in-context learning framework, this longitudinal study addresses this gapby conducting a comprehensive assessment of the robustness of successiveversions of LLMs, vis-\`a-vis GPT-3.5. We conduct extensive experiments toanalyze and understand the impact of the robustness in two distinct learningcategories: zero-shot learning and few-shot learning. Our findings indicatethat, in comparison to earlier versions of LLMs, the updated versions do notexhibit the anticipated level of robustness against adversarial attacks. Inaddition, our study emphasizes the increased effectiveness of synergizedadversarial queries in most zero-shot learning and few-shot learning cases. Wehope that our study can lead to a more refined assessment of the robustness ofLLMs over time and provide valuable insights of these models for bothdevelopers and users.

Robustness Over Time: Understanding Adversarial Examples' Effectiveness on Longitudinal Versions of Large Language Models