analiza aiqbench
Reflex-Guard addresses the critical latency bottleneck in real-time LLM safety filtering by shifting from heavy external APIs or generative judges to a lightweight local model using dense semantic embeddings. By combining jailbreak-aware preprocessing with seven fast binary classifiers, it achieves sub-40ms detection speeds while maintaining high recall against diverse attack vectors like GCG suffixes and Base64 encoding. This approach offers a viable path for deploying robust safety guardrails in latency-sensitive applications without compromising data privacy or response time.
Reflex-Guard: A Low-Latency Guardrail for LLM Prompt Safety Using Dense Semantic Embeddings
arXiv:2608.17556v1 Announce Type: cross Abstract: Large Language Models (LLMs) in real-world applications often face the risks of specially crafted prompts designed to bypass the safety controls. Existing guardrail methods, such as LLM-as-a-judge and cloud-based safety APIs are able to detect unsafe content. However, they often add a delay of about 250-900 ms to each request. This delay is too high for real-time applications, when the system usually needs to respond in less than 100 ms. Furthermore, routing user prompts through external moderation endpoints raises significant data privacy concerns. This paper introduces Reflex-Guard, a lightweight guardrail that runs locally. It uses jailbreak-aware preprocessing, compact sentence-transformer embeddings, and seven fast binary classifiers. Together, these components enable high-accuracy prompt safety filtering with much lower latency than existing solutions. Through systematic evaluation on a strategically balanced dataset of 30,568 samples drawn from five complementary sources, we demonstrate that Reflex-Guard achieves 95.9% recall on harmful prompts at 37.6 ms end-to-end latency. It is faster than existing baselines, including Llama Guard 2 at 255 ms and SafeDecoding at 723 ms. It can detect 100% of GCG suffix attacks and Base64-encoded prompts using the default threshold. However, DrAttack structured prompts required lowering the threshold to 0.03 for optimal detection, as they produced a distinct probability distribution. Reflex-Guard achieves Reflex Efficiency Score (RES) scores up to 16.79, significantly outperforming Llama Guard 2 (11.90) and SafeDecoding (9.80). This analysis offers practical deployment advice and shows that different attack types occupy distinct regions in the embedding probability space.
Abstract:Large Language Models (LLMs) in real-world applications often face the risks of specially crafted prompts designed to bypass the safety controls. Existing guardrail methods, such as LLM-as-a-judge and cloud-based safety APIs are able to detect unsafe content. However, they often add a delay of about 250-900 ms to each request. This delay is too high for real-time applications, when the system usually needs to respond in less than 100 ms. Furthermore, routing user prompts through external moderation endpoints raises significant data privacy concerns. This paper introduces Reflex-Guard, a lightweight guardrail that runs locally. It uses jailbreak-aware preprocessing, compact sentence-transformer embeddings, and seven fast binary classifiers. Together, these components enable high-accuracy prompt safety filtering with much lower latency than existing solutions. Through systematic evaluation on a strategically balanced dataset of 30,568 samples drawn from five complementary sources, we demonstrate that Reflex-Guard achieves 95.9% recall on harmful prompts at 37.6 ms end-to-end latency. It is faster than existing baselines, including Llama Guard 2 at 255 ms and SafeDecoding at 723 ms. It can detect 100% of GCG suffix attacks and Base64-encoded prompts using the default threshold. However, DrAttack structured prompts required lowering the threshold to 0.03 for optimal detection, as they produced a distinct probability distribution. Reflex-Guard achieves Reflex Efficiency Score (RES) scores up to 16.79, significantly outperforming Llama Guard 2 (11.90) and SafeDecoding (9.80). This analysis offers practical deployment advice and shows that different attack types occupy distinct regions in the embedding probability space.
Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2608.17556 [cs.CR]
(or arXiv:2608.17556v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2608.17556 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 345 cuvinte
Subiecte conexe
Building trust in AI agents: A production-ready evaluation framework with Amazon Bedrock AgentCore
AWS · 16 oct. 2026
Presentation: Engineering AI for Creativity and Curiosity on Mobile
InfoQ · 21 iul. 2026
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · 4 sept. 2026
Nemotron 3.5 Content Safety: Customizable Multimodal Safety for Global Enterprise AI
Hugging Face · 4 iun. 2026