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

The rise in malicious usage of large language models, such as fake contentcreation and academic plagiarism, has motivated the development of approachesthat identify AI-generated text, including those based on watermarking oroutlier detection. However, the robustness of these detection algorithms toparaphrases of AI-generated text remains unclear. To stress test thesedetectors, we build a 11B parameter paraphrase generation model (DIPPER) thatcan paraphrase paragraphs, condition on surrounding context, and controllexical diversity and content reordering. Using DIPPER to paraphrase textgenerated by three large language models (including GPT3.5-davinci-003)successfully evades several detectors, including watermarking, GPTZero,DetectGPT, and OpenAI's text classifier. For example, DIPPER drops detectionaccuracy of DetectGPT from 70.3% to 4.6% (at a constant false positive rate of1%), without appreciably modifying the input semantics.  To increase the robustness of AI-generated text detection to paraphraseattacks, we introduce a simple defense that relies on retrievingsemantically-similar generations and must be maintained by a language model APIprovider. Given a candidate text, our algorithm searches a database ofsequences previously generated by the API, looking for sequences that match thecandidate text within a certain threshold. We empirically verify our defenseusing a database of 15M generations from a fine-tuned T5-XXL model and findthat it can detect 80% to 97% of paraphrased generations across differentsettings while only classifying 1% of human-written sequences as AI-generated.We open-source our models, code and data.

Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense