LATIDIA · Ciberseguridad

Decodificación especulativa segura para modelos de lenguaje grandes

arXiv:2610.08678v1 Tipo de anuncio: nuevo Resumen: La decodificación especulativa acelera la inferencia para un modelo de lenguaje grande (LLM), denominado\enf{modelo objetivo}, utilizando primero un modelo más pequeño, denominado\emp

La noticia

arXiv:2610.08678v1 Announce Type: new Abstract: Speculative decoding accelerates inference for a large language model (LLM), referred to as the \emph{target model}, by first using a smaller model, referred to as the \emph{draft model}, to generate candidate tokens and then verifying them with the target model for acceptance or rejection. Prior studies primarily focused on the efficiency-utility trade-off of speculative decoding, e.g., lossy speculative decoding, leaving its security implications largely unexplored. In this work, we bridge this gap by providing the \emph{first} systematic study of the security implications of speculative decoding. Through a large-scale measurement study, we reveal a pronounced security-utility asymmetry: across a wide range of lossy speculative decoding methods,