LATIDIA · Ciberseguridad

ReSI: Mejora recursiva de la seguridad hacia una IA resistente y resiliente

arXiv:2610.12233v1 Tipo de anuncio: nuevo Resumen: La auto-mejora recursiva, la participación de los sistemas de IA en la mejora de sus propias capacidades, está comenzando a pasar de la perspectiva teórica a la práctica, planteando ambas

La noticia

arXiv:2610.12233v1 Announce Type: new Abstract: Recursive self-improvement, the participation of AI systems in improving their own capabilities, is beginning to move from theoretical prospect to practice, posing both challenges and opportunities for safety alignment. Models evolve through frequent updates, and their safety alignment requires continual adaptation to each new checkpoint. Meanwhile, with evolving red-teaming methods exposing new vulnerabilities, safety improvement for each checkpoint needs to mitigate exposed vulnerabilities and generalize to risks not yet revealed. Following R$^2$AI, we term these goals resistance to known threats and resilience to unforeseen risks. Recursive self-improvement, in turn, inspires an approach to both goals: safety alignment could likewise advance through successive rounds of evaluation