LATIDIA · Ciberseguridad

MemForm: Evaluación comparativa del envenenamiento de la formación de memoria en agentes LLM

arXiv:2610.14431v2 Anuncio Tipo: reemplazar Resumen: La memoria a largo plazo permite a los agentes LLM aprender de interacciones pasadas, pero expone la formación de la memoria a la influencia adversaria. Los estudios existentes organizan principalmente la memoria po

La noticia

arXiv:2610.14431v2 Announce Type: replace Abstract: Long-term memory enables LLM agents to learn from past interactions but exposes memory formation to adversarial influence. Existing studies primarily organize memory poisoning by attack channels, corruption patterns, or lifecycle stages, leaving formation interfaces insufficiently characterized. We formulate Memory Formation Poisoning as adversarial manipulation of the interfaces through which persistent agent memory is produced and committed, and introduce MemForm, a controlled diagnostic benchmark covering five interfaces: Content, Evidence, Scope, Selection, and Path Poisoning. We introduce Selection and Path Poisoning as controlled intervention categories that isolate retention and consolidation order effects, respectively. Each category pairs clean and poisoned conditions with interface-specific controls. Evaluations across six LLMs