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

As large language models (LLMs) generate texts with increasing fluency andrealism, there is a growing need to identify the source of texts to prevent theabuse of LLMs. Text watermarking techniques have proven reliable indistinguishing whether a text is generated by LLMs by injecting hidden patternsinto the generated texts. However, we argue that existing watermarking methodsfor LLMs are encoding-inefficient (only contain one bit of information -whether it is generated from an LLM or not) and cannot flexibly meet thediverse information encoding needs (such as encoding model version, generationtime, user id, etc.) in different LLMs application scenarios. In this work, weconduct the first systematic study on the topic of Codable Text Watermarkingfor LLMs (CTWL) that allows text watermarks to carry more customizableinformation. First of all, we study the taxonomy of LLM watermarking technologyand give a mathematical formulation for CTWL. Additionally, we provide acomprehensive evaluation system for CTWL: (1) watermarking success rate, (2)robustness against various corruptions, (3) coding rate of payload information,(4) encoding and decoding efficiency, (5) impacts on the quality of thegenerated text. To meet the requirements of these non-Pareto-improving metrics,we devise a CTWL method named Balance-Marking, based on the motivation ofensuring that available and unavailable vocabularies for encoding informationhave approximately equivalent probabilities. Compared to the random vocabularypartitioning extended from the existing work, a probability-balanced vocabularypartition can significantly improve the quality of the generated text.Extensive experimental results have shown that our method outperforms a directbaseline under comprehensive evaluation.

Towards Codable Watermarking for Injecting Multi-bit Information to LLM