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 widespread use of Large Language Models (LLMs), celebrated for theirability to generate human-like text, has raised concerns about misinformationand ethical implications. Addressing these concerns necessitates thedevelopment of robust methods to detect and attribute text generated by LLMs.This paper investigates "Cross-Model Detection," evaluating whether aclassifier trained to distinguish between source LLM-generated andhuman-written text can also detect text from a target LLM without furthertraining. The study comprehensively explores various LLM sizes and families,and assesses the impact of conversational fine-tuning techniques on classifiergeneralization. The research also delves into Model Attribution, encompassingsource model identification, model family classification, and model sizeclassification. Our results reveal several key findings: a clear inverserelationship between classifier effectiveness and model size, with larger LLMsbeing more challenging to detect, especially when the classifier is trained ondata from smaller models. Training on data from similarly sized LLMs canimprove detection performance from larger models but may lead to decreasedperformance when dealing with smaller models. Additionally, model attributionexperiments show promising results in identifying source models and modelfamilies, highlighting detectable signatures in LLM-generated text. Overall,our study contributes valuable insights into the interplay of model size,family, and training data in LLM detection and attribution.

From Text to Source: Results in Detecting Large Language Model-Generated Content