LATIDIA · Ciberseguridad

Análisis de la vulnerabilidad de LLM a errores transitorios de GPU: un estudio de inyección de fallos a nivel de instrucción.

arXiv: 2601.19912v2Announce Type: replace-cross Resumen: Los modelos de lenguaje grandes (LLM) son altamente intensivos en cómputo y memoria, lo que plantea demandas significativas en las GPU de alto rendimiento. Al mismo tiempo, los avances en la tecnología de GPU

La noticia

arXiv:2601.19912v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are highly compute- and memory-intensive, posing significant demands on high-performance GPUs. At the same time, advances in GPU technology driven by shrinking transistor sizes and lower operating voltages have made these devices increasingly susceptible to soft errors, particularly in radiation-prone environments such as satellite computing scenarios. While prior work has examined GPU soft errors, most studies have focused on general-purpose applications or conventional neural networks primarily used for vision tasks such as classification and detection. However, vulnerability analysis of modern LLMs remains limited, despite their rapid adoption in diverse application scenarios. Given the unique characteristics of LLMs, their resilience to soft