Prompt Overflow: What the Guardrail Inspects Is Not What the Model Infers

Abstract

Guardrail models (a.k.a. safety checkers) are widely deployed to screen user inputs before they reach large language models (LLMs), serving as a primary defense against prompt injection attacks. Due to strict context constraints, these models handle overlength prompts through truncation or segmentation-based inspection. While prior work has focused on semantic adversarial inputs, the security implications of these long-input processing mechanisms remain largely unexplored. In this paper, we iden

Categories

Framework mappings

Suggested from the entry's categories.

@misc{zhou2026prompt,
  title = {{Prompt Overflow: What the Guardrail Inspects Is Not What the Model Infers}},
  author = {Yuanbo Zhou and Changjia Zhu and Junyu Wang and Xu He and Yan Zhai and Kun Sun and Mingkui Wei and Junjie Xiong},
  year = {2026},
  month = may,
  eprint = {2605.23196},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.23196}
}