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 been widely used in various applicationsbut are known to suffer from issues related to untruthfulness and toxicity.While parameter-efficient modules (PEMs) have demonstrated their effectivenessin equipping models with new skills, leveraging PEMs for deficiency unlearningremains underexplored. In this work, we propose a PEMs operation approach,namely Extraction-before-Subtraction (Ext-Sub), to enhance the truthfulness anddetoxification of LLMs through the integration of ``expert'' PEM and``anti-expert'' PEM. Remarkably, even anti-expert PEM possess valuablecapabilities due to their proficiency in generating fabricated content, whichnecessitates language modeling and logical narrative competence. Rather thanmerely negating the parameters, our approach involves extracting andeliminating solely the deficiency capability within anti-expert PEM whilepreserving the general capabilities. To evaluate the effectiveness of ourapproach in terms of truthfulness and detoxification, we conduct extensiveexperiments on LLMs, encompassing additional abilities such as languagemodeling and mathematical reasoning. Our empirical results demonstrate that ourapproach effectively improves truthfulness and detoxification, while largelypreserving the fundamental abilities of LLMs.

Separate the Wheat from the Chaff: Model Deficiency Unlearning via Parameter-Efficient Module Operation