Metraipolis › archive › dc_c8efe5180e19
RAISED: Self-Distillation for Robustness to Prompt Injection in LLM Agents — arXiv:2610.06401
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.06401
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T05:27:38.089Z
- SHA-256 of the archived text
d7666e64e463af08cea81337329b64835f05e582d19562621cab2322f91c1a0c- Stable URL of this page
https://metraipolis.com/documento/dc_c8efe5180e19
How to cite
RAISED: Self-Distillation for Robustness to Prompt Injection in LLM Agents — arXiv:2610.06401 — arXiv, https://arxiv.org/abs/2610.06401 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c8efe5180e19 (Metraipolis archive dc_c8efe5180e19, sha256 d7666e64e463af08cea81337329b64835f05e582d19562621cab2322f91c1a0c)
BibTeX
@misc{dcc8efe5180e19,
title = {RAISED: Self-Distillation for Robustness to Prompt Injection in LLM Agents — arXiv:2610.06401},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.06401},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c8efe5180e19; sha256 d7666e64e463af08cea81337329b64835f05e582d19562621cab2322f91c1a0c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.06401. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Tool-using language-model agents are vulnerable to indirect prompt injection because they must act on untrusted external content. Existing training-time defenses can reduce attack success rates, but often at the cost of general capabilities. We show that training-based defenses induce substantial drift in the model's output distribution, altering its behavior even in benign settings and providing a potential mechanism for utility degradation. We further identify a failure mode of these defenses: On benign tool-use tasks, the model refrains from a step needed to finish an authorized task, particularly when that step is indicated by a tool output. To address these limitations, we introduce RAISED (Robust Attack Invariance through Self-Distillation), a training framework that combines self-generation and self-distillation. The model first generates its own tool-use scenarios, with an emphasis on cases where task completion requires acting on legitimate guidance from tool outputs. Then, through self-distillation, the student is trained to match the teacher's clean-context behavior on both clean and injected variants of the same trajectory. RAISED substantially reduces the attack success rate of prompt injections in tool responses while, unlike prior training-based defenses, preserving utility on both agentic and general-purpose benchmarks. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.06401 — 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 2026-10-06T05:24:06.150Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).