Enregistré dans:
| Auteurs principaux: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2504.20965 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866909648908451840 |
|---|---|
| author | Cai, Zikui Shabihi, Shayan An, Bang Che, Zora Bartoldson, Brian R. Kailkhura, Bhavya Goldstein, Tom Huang, Furong |
| author_facet | Cai, Zikui Shabihi, Shayan An, Bang Che, Zora Bartoldson, Brian R. Kailkhura, Bhavya Goldstein, Tom Huang, Furong |
| contents | We introduce AegisLLM, a cooperative multi-agent defense against adversarial attacks and information leakage. In AegisLLM, a structured workflow of autonomous agents - orchestrator, deflector, responder, and evaluator - collaborate to ensure safe and compliant LLM outputs, while self-improving over time through prompt optimization. We show that scaling agentic reasoning system at test-time - both by incorporating additional agent roles and by leveraging automated prompt optimization (such as DSPy)- substantially enhances robustness without compromising model utility. This test-time defense enables real-time adaptability to evolving attacks, without requiring model retraining. Comprehensive evaluations across key threat scenarios, including unlearning and jailbreaking, demonstrate the effectiveness of AegisLLM. On the WMDP unlearning benchmark, AegisLLM achieves near-perfect unlearning with only 20 training examples and fewer than 300 LM calls. For jailbreaking benchmarks, we achieve 51% improvement compared to the base model on StrongReject, with false refusal rates of only 7.9% on PHTest compared to 18-55% for comparable methods. Our results highlight the advantages of adaptive, agentic reasoning over static defenses, establishing AegisLLM as a strong runtime alternative to traditional approaches based on model modifications. Code is available at https://github.com/zikuicai/aegisllm |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_20965 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Security Cai, Zikui Shabihi, Shayan An, Bang Che, Zora Bartoldson, Brian R. Kailkhura, Bhavya Goldstein, Tom Huang, Furong Machine Learning We introduce AegisLLM, a cooperative multi-agent defense against adversarial attacks and information leakage. In AegisLLM, a structured workflow of autonomous agents - orchestrator, deflector, responder, and evaluator - collaborate to ensure safe and compliant LLM outputs, while self-improving over time through prompt optimization. We show that scaling agentic reasoning system at test-time - both by incorporating additional agent roles and by leveraging automated prompt optimization (such as DSPy)- substantially enhances robustness without compromising model utility. This test-time defense enables real-time adaptability to evolving attacks, without requiring model retraining. Comprehensive evaluations across key threat scenarios, including unlearning and jailbreaking, demonstrate the effectiveness of AegisLLM. On the WMDP unlearning benchmark, AegisLLM achieves near-perfect unlearning with only 20 training examples and fewer than 300 LM calls. For jailbreaking benchmarks, we achieve 51% improvement compared to the base model on StrongReject, with false refusal rates of only 7.9% on PHTest compared to 18-55% for comparable methods. Our results highlight the advantages of adaptive, agentic reasoning over static defenses, establishing AegisLLM as a strong runtime alternative to traditional approaches based on model modifications. Code is available at https://github.com/zikuicai/aegisllm |
| title | AegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Security |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2504.20965 |