AI Security Research Portal
Sourcessourceseed2026-07-04ai-securitymemory-auditcausal-attributionanomaly-detectioncounterfactual-replaypoisoning

MemAudit

Collection Summary

Harmful behavior가 관측된 뒤 어떤 stored memory가 원인이었는지 counterfactual replay와 memory consistency graph로 추적하는 post-hoc auditing framework다.

Rollout-Buffer Relevance