![[2602.16935] DeepContext: Stateful Real-Time Detection of Multi-Turn Adversarial Intent Drift in LLMs](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.16935] DeepContext: Stateful Real-Time Detection of Multi-Turn Adversarial Intent Drift in LLMs
Summary
The paper introduces DeepContext, a stateful framework for detecting adversarial intent drift in multi-turn dialogues within large language models, significantly improving detection accuracy and efficiency.
Why It Matters
As large language models become more prevalent, ensuring their safety against adversarial attacks is critical. DeepContext addresses the limitations of stateless models by providing a method to track user intent over multiple interactions, enhancing the security of AI systems.
Key Takeaways
- DeepContext utilizes a Recurrent Neural Network to track user intent over multiple dialogue turns.
- The framework achieves a state-of-the-art F1 score of 0.84 in detecting adversarial intent drift.
- DeepContext offers a computationally efficient solution with sub-20ms inference time on a T4 GPU.
- The model addresses the 'Safety Gap' in existing stateless systems, improving real-time application viability.
- This research highlights the importance of temporal awareness in enhancing AI safety measures.
Computer Science > Artificial Intelligence arXiv:2602.16935 (cs) [Submitted on 18 Feb 2026] Title:DeepContext: Stateful Real-Time Detection of Multi-Turn Adversarial Intent Drift in LLMs Authors:Justin Albrethsen, Yash Datta, Kunal Kumar, Sharath Rajasekar View a PDF of the paper titled DeepContext: Stateful Real-Time Detection of Multi-Turn Adversarial Intent Drift in LLMs, by Justin Albrethsen and 3 other authors View PDF HTML (experimental) Abstract: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...