Who's Behind the Text? Robust Model Authorship Attribution

Key points

• As large language models (LLMs) are increasingly used to generate text, their misuse has become a growing societal problem, with implications for academic misconduct, model distillation, and cyberattacks.1
• For these reasons, Model Authorship Attribution (MAA), which identifies the source model of a given text, is important for promoting the responsible use of LLMs and improving AI safety.1
• ATLAS learns how text from each source family is generated and attributes a given text by comparing its likelihood under each learned source condition.1
• We further introduce MAABench, a broad MAA benchmark for evaluating practical OOD settings such as domain and version.1

Sources (1)

• [1]Who's Behind the Text? Robust Model Authorship Attribution
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 04:54 AM

  To address this challenge, we propose Authorship Tracing via Likelihood-based Attribution of Sources (ATLAS), which formulates MAA through source-conditioned text modeling rather than direct source prediction.

  As large language models (LLMs) are increasingly used to generate text, their misuse has become a growing societal problem, with implications for academic misconduct, model distillation, and cyberattacks.

Extractive summary: sentences quoted from the sources.