Metraipolis › archive › dc_0cdae00b4f74

ShieldCLIP: Selective Safety Alignment for Harmful Content Mitigation in Multimodal Foundation Models — arXiv:2609.39688

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39688
License (as archived)
CC0-1.0
Archived
2026-10-07T12:45:42.093Z
SHA-256 of the archived text
3dcf340d623b2d03c051cae4fbf0c57ce1ae6646f0e9706330a92dd56ad164db
Stable URL of this page
https://metraipolis.com/documento/dc_0cdae00b4f74

How to cite

ShieldCLIP: Selective Safety Alignment for Harmful Content Mitigation in Multimodal Foundation Models — arXiv:2609.39688 — arXiv (en), https://arxiv.org/abs/2609.39688 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0cdae00b4f74 (Metraipolis archive dc_0cdae00b4f74, sha256 3dcf340d623b2d03c051cae4fbf0c57ce1ae6646f0e9706330a92dd56ad164db)
BibTeX
@misc{dc0cdae00b4f74,
  title = {ShieldCLIP: Selective Safety Alignment for Harmful Content Mitigation in Multimodal Foundation Models — arXiv:2609.39688},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39688},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0cdae00b4f74; sha256 3dcf340d623b2d03c051cae4fbf0c57ce1ae6646f0e9706330a92dd56ad164db},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39688. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multimodal encoders such as CLIP underlie many downstream systems, but their web-scale training data embed harmful associations that safety alignment must suppress without unnecessarily changing benign representations. Because ethical and practical constraints prevent collecting real unsafe content at scale, existing datasets pair safe real samples with generated counterparts, but label every generated sample unsafe, even when one modality is individually safe. To address this, we introduce ShieldCLIP, the first framework to condition safety alignment on the observed safety state of each modality rather than the origin of a sample, preserving safe content while redirecting only what is unsafe. We also introduce ViSUv2, a 195k-quadruplet dataset with independent per-modality safety labels across 578 concepts and 28 categories. Using these labels, ShieldCLIP defines a four-way conditional objective beyond pair-level supervision: safe content is anchored, unsafe modalities are redirected to their safe counterparts, mixed pairs update only the unsafe branch, and coherence is enforced when both are unsafe. We evaluate ShieldCLIP on cross-modal retrieval, text-to-image generation with Stable Diffusion v1.4 and SDXL, and image-to-text generation with LLaVA. Across these settings, ShieldCLIP consistently reduces harmful outputs over prior safety-aligned encoders and strong mitigation baselines, while preserving the utility of the original embedding space. Extensive ablation studies further show that both modality-specific supervision and the selective alignment objective contribute to these gains. Source code, trained models, and ViSUv2 (under a controlled-access protocol) will be made publicly available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multimedia (cs.MM).
Pagina del lavoro: https://arxiv.org/abs/2609.39688 — 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 2026-10-07T12:14:14.136Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).