AI RESEARCH
TwinGate: Stateful Defense against Decompositional Jailbreaks in Untraceable Traffic via Asymmetric Contrastive Learning
arXiv CS.LG
•
ArXi:2604.27861v1 Announce Type: cross Decompositional jailbreaks pose a critical threat to large language models (LLMs) by allowing adversaries to fragment a malicious objective into a sequence of individually benign queries that collectively reconstruct prohibited content. In real-world deployments, LLMs face a continuous, untraceable stream of fully anonymized and arbitrarily interleaved requests, infiltrated by covertly distributed adversarial queries. Under this rigorous threat model, state-of-the-art defensive strategies exhibit fundamental limitations.