Adversarial Distilled Retrieval-Augmented Guarding Model for Online Malicious Intent Detection
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909881495191552 |
|---|---|
| author | Guo, Yihao Bian, Haocheng Zhou, Liutong Wang, Ze Zhang, Zhaoyi Kawala, Francois Dean, Milan Fischer, Ian Peng, Yuantao Tokgozoglu, Noyan Barrientos, Ivan Shaik, Riyaaz Li, Rachel Venkataraman, Chandru Far, Reza Shifteh Pawar, Moses Sundaranatha, Venkat Xu, Michael Chu, Frank |
| author_facet | Guo, Yihao Bian, Haocheng Zhou, Liutong Wang, Ze Zhang, Zhaoyi Kawala, Francois Dean, Milan Fischer, Ian Peng, Yuantao Tokgozoglu, Noyan Barrientos, Ivan Shaik, Riyaaz Li, Rachel Venkataraman, Chandru Far, Reza Shifteh Pawar, Moses Sundaranatha, Venkat Xu, Michael Chu, Frank |
| contents | With the deployment of Large Language Models (LLMs) in interactive applications, online malicious intent detection has become increasingly critical. However, existing approaches fall short of handling diverse and complex user queries in real time. To address these challenges, we introduce ADRAG (Adversarial Distilled Retrieval-Augmented Guard), a two-stage framework for robust and efficient online malicious intent detection. In the training stage, a high-capacity teacher model is trained on adversarially perturbed, retrieval-augmented inputs to learn robust decision boundaries over diverse and complex user queries. In the inference stage, a distillation scheduler transfers the teacher's knowledge into a compact student model, with a continually updated knowledge base collected online. At deployment, the compact student model leverages top-K similar safety exemplars retrieved from the online-updated knowledge base to enable both online and real-time malicious query detection. Evaluations across ten safety benchmarks demonstrate that ADRAG, with a 149M-parameter model, achieves 98.5% of WildGuard-7B's performance, surpasses GPT-4 by 3.3% and Llama-Guard-3-8B by 9.5% on out-of-distribution detection, while simultaneously delivering up to 5.6x lower latency at 300 queries per second (QPS) in real-time applications. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_14622 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Adversarial Distilled Retrieval-Augmented Guarding Model for Online Malicious Intent Detection Guo, Yihao Bian, Haocheng Zhou, Liutong Wang, Ze Zhang, Zhaoyi Kawala, Francois Dean, Milan Fischer, Ian Peng, Yuantao Tokgozoglu, Noyan Barrientos, Ivan Shaik, Riyaaz Li, Rachel Venkataraman, Chandru Far, Reza Shifteh Pawar, Moses Sundaranatha, Venkat Xu, Michael Chu, Frank Cryptography and Security Artificial Intelligence Machine Learning With the deployment of Large Language Models (LLMs) in interactive applications, online malicious intent detection has become increasingly critical. However, existing approaches fall short of handling diverse and complex user queries in real time. To address these challenges, we introduce ADRAG (Adversarial Distilled Retrieval-Augmented Guard), a two-stage framework for robust and efficient online malicious intent detection. In the training stage, a high-capacity teacher model is trained on adversarially perturbed, retrieval-augmented inputs to learn robust decision boundaries over diverse and complex user queries. In the inference stage, a distillation scheduler transfers the teacher's knowledge into a compact student model, with a continually updated knowledge base collected online. At deployment, the compact student model leverages top-K similar safety exemplars retrieved from the online-updated knowledge base to enable both online and real-time malicious query detection. Evaluations across ten safety benchmarks demonstrate that ADRAG, with a 149M-parameter model, achieves 98.5% of WildGuard-7B's performance, surpasses GPT-4 by 3.3% and Llama-Guard-3-8B by 9.5% on out-of-distribution detection, while simultaneously delivering up to 5.6x lower latency at 300 queries per second (QPS) in real-time applications. |
| title | Adversarial Distilled Retrieval-Augmented Guarding Model for Online Malicious Intent Detection |
| topic | Cryptography and Security Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2509.14622 |