2601.10313v1.pdf

{
  "dorsal/arxiv": [
    {
      "date_modified": "2026-02-17T05:53:23.836000Z",
      "hash": "32cd4747bf9546986d33668f913461ed534bbe4c9bb64a94240aa3913c5a4588",
      "id": "a5d5d0fb-2069-4f9b-ab8c-f09677d9d634",
      "source": {
        "execution_id": "ddf9899b-f337-48e9-bcbb-8ec99d8be428",
        "id": "arXiv Dataset",
        "type": "Model",
        "variant": "snapshot-2026-01-17",
        "version": "0.1.0"
      },
      "user_id": 1000002
    }
  ],
  "file/base": {
    "record": {
      "extension": ".pdf",
      "hash": "32cd4747bf9546986d33668f913461ed534bbe4c9bb64a94240aa3913c5a4588",
      "media_type": "application/pdf",
      "media_type_prefix": "application",
      "name": "2601.10313v1.pdf",
      "size": 1758478
    },
    "source": {
      "execution_id": "e4b6eaad-427c-49eb-a913-824287e5f79a",
      "id": "file/base",
      "type": "Model",
      "version": "1.0.0"
    }
  },
  "file/pdf": {
    "private": false,
    "record": {
      "keywords": [],
      "page_count": 15,
      "producer": "pikepdf 8.15.1",
      "title": "Hierarchical Refinement of Universal Multimodal Attacks on Vision-Language Models",
      "version": "1.7"
    },
    "source": {
      "execution_id": "c4fa3820-a985-48a4-a72b-63f24af8fce6",
      "id": "dorsal/pdf",
      "type": "Model",
      "version": "1.1.0"
    }
  }
}