Metraipolis › archive › dc_67bc89f77a4c
PGuard: A Native, Streaming, and Plug-in Guardrail for Multimodal Large Language Models — arXiv:2611.02791
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.02791
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T11:10:54.915Z
- SHA-256 of the archived text
d34b6cfc8cb80dd006e4ad964ec299fe84f5c86cb14b99bd6c9debff19589a79- Stable URL of this page
https://metraipolis.com/documento/dc_67bc89f77a4c
How to cite
PGuard: A Native, Streaming, and Plug-in Guardrail for Multimodal Large Language Models — arXiv:2611.02791 — arXiv (en), https://arxiv.org/abs/2611.02791 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_67bc89f77a4c (Metraipolis archive dc_67bc89f77a4c, sha256 d34b6cfc8cb80dd006e4ad964ec299fe84f5c86cb14b99bd6c9debff19589a79)
BibTeX
@misc{dc67bc89f77a4c,
title = {PGuard: A Native, Streaming, and Plug-in Guardrail for Multimodal Large Language Models — arXiv:2611.02791},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.02791},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_67bc89f77a4c; sha256 d34b6cfc8cb80dd006e4ad964ec299fe84f5c86cb14b99bd6c9debff19589a79},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.02791. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
As streaming generation becomes the dominant serving paradigm for both text-only and multimodal large language models, ensuring safety throughout the generation process has become a critical challenge. We introduce PGuard, a native, streaming, and plug-in guardrail that transforms safety monitoring from an external serving dependency into an integrated capability of the backbone model. PGuard directly leverages the backbone's intermediate representations for token-level safety detection, without altering its original inference pathway or generation behavior. Specifically, dynamic layer gating selectively aggregates residual-stream representations across layers to extract safety-relevant signals during both prefill and autoregressive decoding. To balance detection accuracy and computational efficiency, PGuard supports multiple representation-modeling architectures, including a lightweight stateless variant and sequence-aware variants built on multi-scale causal EMA-attention and GRU-based recurrent modules. Its design also accounts for the execution characteristics of LLM serving frameworks, such as SGLang, optimizing computation and data access to minimize additional runtime and synchronization overhead. As a result, PGuard incurs only 0.82-2.26% end-to-end overhead on the hundred-billion-parameter backbone Qwen3.5-122B-A10B served by SGLang. Extensive evaluations across more than 30 benchmarks spanning text and multimodal safety, jailbreaks, cyber and CBRN risks, multilingual scenarios, and real-world traffic demonstrate state-of-the-art performance among native guardrails. On text queries with Qwen3.5-9B, PGuard achieves an average F1 score of 0.836, approaching dedicated external guardrails, while supporting deployment-time configuration of category-specific thresholds and enforcement actions. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2611.02791 — 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 2026-11-03T10:59:52.351Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).