LATIDIA · Ciberseguridad
Marcas de agua robustas «plug-and-play» con múltiples licenciatarios para LLM de mezcla de expertos
arXiv:2611.23838v1 Tipo de anuncio: nuevo Resumen: Las arquitecturas de mezcla de expertos (MoE) permiten crear modelos de lenguaje grandes (LLM) de alta capacidad con un alto nivel de computación dispersa, lo que las hace atractivas para su implementación en hardware privado.
WhatsApp ↗Telegram ↗
La noticia
arXiv:2611.23838v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) architectures enable high-capacity large language models (LLMs) with high sparse computation, making them attractive for deployment on private hardware. However, distributing model weights exposes valuable intellectual property (IP) to leakage and unauthorized redistribution, creating an urgent need to both verify model ownership and trace leaked copies to individual licensees. Existing model-watermarking techniques are ill-suited for meeting this need because they are fragile under MoE's dynamic routing or require costly retraining to embed a distinct watermark for each licensee, thus making scalable distribution difficult. To address these challenges, we present PPMoE, the first plug-and-play multi-licensee watermarking framework for MoE LLMs. Rather than training a