LATIDIA · Ciberseguridad

SaplingGuard: una barandilla multiagente consciente del perfil multidimensional para la interacción adolescente-LLM segura para el desarrollo

arXiv:2609.35892v1 Anuncio Tipo: nuevo Resumen: A medida que los adolescentes utilizan cada vez más los LLM en la vida cotidiana, garantizar respuestas seguras y apropiadas para el desarrollo se ha vuelto esencial. Sin embargo, las medidas de protección de los LLM existentes son fundamentales.

La noticia

arXiv:2609.35892v1 Announce Type: new Abstract: As adolescents increasingly use LLMs in everyday life, ensuring safe and developmentally appropriate responses has become essential. However, existing LLM guardrails primarily target explicit harmful content in isolated prompts or responses and are less effective at identifying implicit, context-dependent developmental risks. To address this limitation, we propose SaplingGuard, a plug-and-play, profile-aware and dialogue-aware guardrail that requires no modification to downstream model parameters. SaplingGuard decomposes adolescent safety intervention into three specialized agents for user profile construction, context-aware risk assessment, and intent-preserving prompt optimization. Together, these agents leverage the current prompt, preceding dialogue, and structured user characteristics to identify contextual risks and guide downstream response generation. We