2026/02/18 by Justin Albrethsen, Yash Datta, Kunal Kumar +1 · 1 voice
Computer Science · #Adversarial Robustness in Machine Learning #Explainable Artificial Intelligence (XAI) #Inference #Overhead (engineering) #Pruning #Recurrent neural network #Sequence (biology) #State (computer science) #Stateful firewall #Stateless protocol #Topic Modeling #Trajectory #cs.AI #cs.ET #cs.LG
paper · pdf · doi:10.48550/arxiv.2602.16935
openalex publication_date 2026/02/18 · arxiv published 2026/02/18 · arxiv updated 2026/02/18 · openalex created_date 2026/02/21 · openalex updated_date 2026/07/28
While Large Language Model (LLM) capabilities have scaled, safety guardrails remain largely stateless, treating multi-turn dialogues as a series of disconnected events. This lack of temporal awareness facilitates a "Safety Gap" where adversarial tactics, like Crescendo and ActorAttack, slowly bleed malicious intent across turn boundaries to bypass stateless filters. We introduce DeepContext, a stateful monitoring framework designed to map the temporal trajectory of user intent. DeepContext discards the isolated evaluation model in favor of a Recurrent Neural Network (RNN) architecture that ingests a sequence of fine-tuned turn-level embeddings. By propagating a hidden state across the conversation, DeepContext captures the incremental accumulation of risk that stateless models overlook. Our evaluation demonstrates that DeepContext significantly outperforms existing baselines in multi-turn jailbreak detection, achieving a state-of-the-art F1 score of 0.84, which represents a substantial improvement over both hyperscaler cloud-provider guardrails and leading open-weight models such as Llama-Prompt-Guard-2 (0.67) and Granite-Guardian (0.67). Furthermore, DeepContext maintains a sub-20ms inference overhead on a T4 GPU, ensuring viability for real-time applications. These results suggest that modeling the sequential evolution of intent is a more effective and computationally efficient alternative to deploying massive, stateless models.