Metraipolis › archive › dc_2944511366a4

Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMs — arXiv:2609.26185

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.26185
License (as archived)
CC0-1.0
Archived
2026-10-08T13:47:04.946Z
SHA-256 of the archived text
13cfa6ddc190a1110c428f6257c3921631aaa41b2d8c7b3d26ddd04662a24c66
Stable URL of this page
https://metraipolis.com/documento/dc_2944511366a4

How to cite

Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMs — arXiv:2609.26185 — arXiv (en), https://arxiv.org/abs/2609.26185 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2944511366a4 (Metraipolis archive dc_2944511366a4, sha256 13cfa6ddc190a1110c428f6257c3921631aaa41b2d8c7b3d26ddd04662a24c66)
BibTeX
@misc{dc2944511366a4,
  title = {Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMs — arXiv:2609.26185},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.26185},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2944511366a4; sha256 13cfa6ddc190a1110c428f6257c3921631aaa41b2d8c7b3d26ddd04662a24c66},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26185. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large Language Models (LLMs) demonstrate impressive capabilities across many applications but remain vulnerable to jailbreak attacks, which elicit harmful or unintended content. While model fine-tuning is an option for safety alignment, it is costly and prone to catastrophic forgetting. Prompt optimization has emerged as a promising alternative, yet existing prompt-based defenses typically rely on static modifications (e.g., fixed prefixes or suffixes) that cannot adapt to diverse and evolving attacks. We propose Dynamic Deep Prompt Optimization (DDPO), the first jailbreak defense based on deep prompt optimization. DDPO uses the target LLM's own intermediate layers as feature extractors to dynamically generate defensive embeddings via a lightweight multilayer perceptron. These tailored embeddings are then injected into a subsequent intermediate layer, enabling an input-dependent defense without modifying the LLM's weights. This design ensures high adaptability with minimal computational overhead. Experiments on a diverse set of models and attacks demonstrate that DDPO significantly outperforms static prompt optimization methods, particularly on weakly aligned models and when handling semantically ambiguous benign prompts, successfully distinguishing them from genuinely harmful requests.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.AI).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.26185 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T13:18:38.294Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).