Metraipolis › archive › dc_e0912499fb82

Your Agentic LLMs Secretly Encode Indirect Prompt-Injection Exposure in Hidden States — arXiv:2608.02657

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.02657
License (as archived)
CC0-1.0
Archived
2026-10-31T12:17:31.918Z
SHA-256 of the archived text
831e4da6a10faef613d70c4931897284503265c5aae6ecd6c73e5c43e15d58f2
Stable URL of this page
https://metraipolis.com/documento/dc_e0912499fb82

How to cite

Your Agentic LLMs Secretly Encode Indirect Prompt-Injection Exposure in Hidden States — arXiv:2608.02657 — arXiv (en), https://arxiv.org/abs/2608.02657 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e0912499fb82 (Metraipolis archive dc_e0912499fb82, sha256 831e4da6a10faef613d70c4931897284503265c5aae6ecd6c73e5c43e15d58f2)
BibTeX
@misc{dce0912499fb82,
  title = {Your Agentic LLMs Secretly Encode Indirect Prompt-Injection Exposure in Hidden States — arXiv:2608.02657},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.02657},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e0912499fb82; sha256 831e4da6a10faef613d70c4931897284503265c5aae6ecd6c73e5c43e15d58f2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.02657. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Agentic LLMs are vulnerable to indirect prompt injection (IPI) attacks, e.g., malicious side-tasks hidden in external tool results. While many efforts have sought to address this threat, little is known about the internals of agentic LLMs when they are exposed to IPI attacks. For simplicity, we refer to this condition as IPI exposure. In this paper, we study IPI exposure from three perspectives. (1) Probing: Across eight models, including the 753B-parameter GLM-5.2 and the 2.8T-parameter Kimi-K3, simple linear probes trained on pre-generation hidden states can predict LLMs' IPI exposure. These probes achieve 0.90+ AUROC on unseen attacks, agent instructions, and task suites; they remain robustly predictive under adaptive attacks and in cross-lingual settings. (2) Defense: We reveal and diagnose a knowledge-action gap: post-trained LLMs encode signals predictive of IPI exposure, yet do not reliably bind these signals to safe agentic actions. We therefore introduce a probe-gated reasoning-based defense to bridge this gap at test time. On difficult AgentDojo settings, it substantially reduces attack success rate, e.g., from 34.6% to 0% on Qwen3.5-27B, and better preserves clean-task utility than the baselines. (3) Explanation: We introduce an analysis framework that identifies natural-language explanations strongly correlated with probe-captured signals. The resulting profiles differ across models: latent signals can align with either direct IPI-exposure sensing or indirect operational cues. Code is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.AI).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.02657 — 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 2026-10-31T12:05:52.910Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).