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

Language generation models have been an increasingly powerful enabler formany applications. Many such models offer free or affordable API access, whichmakes them potentially vulnerable to model extraction attacks throughdistillation. To protect intellectual property (IP) and ensure fair use ofthese models, various techniques such as lexical watermarking and synonymreplacement have been proposed. However, these methods can be nullified byobvious countermeasures such as "synonym randomization". To address this issue,we propose GINSEW, a novel method to protect text generation models from beingstolen through distillation. The key idea of our method is to inject secretsignals into the probability vector of the decoding steps for each targettoken. We can then detect the secret message by probing a suspect model to tellif it is distilled from the protected one. Experimental results show thatGINSEW can effectively identify instances of IP infringement with minimalimpact on the generation quality of protected APIs. Our method demonstrates anabsolute improvement of 19 to 29 points on mean average precision (mAP) indetecting suspects compared to previous methods against watermark removalattacks.

Protecting Language Generation Models via Invisible Watermarking