Metraipolis › archive › dc_d3d98ccb0068
On Task, Off Track: Benchmarking LLM and Agent Robustness to Near-Target Prompt Injection — arXiv:2610.32531
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.32531
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-26T20:07:26.284Z
- SHA-256 of the archived text
0aaf8d1a17a0b174c22ccdcadd0d7e3c97770f30353cd4859295f46997458725- Stable URL of this page
https://metraipolis.com/documento/dc_d3d98ccb0068
How to cite
On Task, Off Track: Benchmarking LLM and Agent Robustness to Near-Target Prompt Injection — arXiv:2610.32531 — arXiv (en), https://arxiv.org/abs/2610.32531 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d3d98ccb0068 (Metraipolis archive dc_d3d98ccb0068, sha256 0aaf8d1a17a0b174c22ccdcadd0d7e3c97770f30353cd4859295f46997458725)
BibTeX
@misc{dcd3d98ccb0068,
title = {On Task, Off Track: Benchmarking LLM and Agent Robustness to Near-Target Prompt Injection — arXiv:2610.32531},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.32531},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d3d98ccb0068; sha256 0aaf8d1a17a0b174c22ccdcadd0d7e3c97770f30353cd4859295f46997458725},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.32531. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Prompt injection benchmarks assess whether large language models (LLMs) and agents resist malicious instructions embedded in external content. However, a system could perform well by rejecting attacks that obviously abandon the user's request, while still accepting changes that retain most of a legitimate answer but make it incorrect or harmful. Testing more instructions for unrelated outcomes would not necessarily reveal this weakness. We call attacks seeking responses designed to resemble legitimate answers while introducing incorrect or harmful changes near-target prompt injection. We introduce NTBench, the Near-Target Benchmark, which builds on legitimate tasks from existing datasets. For each task, a legitimate reference outcome guides the generation of target--instruction pairs, with separate tracks targeting incorrect and harmful outcomes in responses or agent tool invocations. Harmful targets can also accompany otherwise correct task completion. We combine LLM-based checks with human validation of the intended deviations, then assess whether injected instructions induce them using response judgments or tool-call checks, excluding deviations already observed without injection. We evaluate seven LLMs on ten datasets for language tasks and two agent benchmarks. Under a common budget of three attack executions per sample, both tracks achieve higher overall attack success rates than the fixed-target baseline for every evaluated model in each setting. On language tasks, the increases range from 9.3 to 35.8 percentage points for incorrect targets and 1.7 to 17.3 percentage points for harmful targets. These results show that resistance to the tested fixed-target attacks does not imply resistance to related incorrect or harmful outcomes, and that reference-guided targets broaden the failures revealed by robustness evaluation. Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2610.32531 — 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 2026-10-26T19:48:33.268Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).