2601.05776v1.pdf

{
  "dorsal/arxiv": [
    {
      "date_modified": "2026-02-17T05:53:05.135000Z",
      "hash": "615fdc8a3ee8ef23e27500d86bd982eb2b76ec4920b3c81f284b1bf12464ddb6",
      "id": "815715d8-a995-4033-aba2-c4f425d34da1",
      "source": {
        "execution_id": "73d0406c-45b5-48a2-8d75-b87eb4d29c20",
        "id": "arXiv Dataset",
        "type": "Model",
        "variant": "snapshot-2026-01-17",
        "version": "0.1.0"
      },
      "user_id": 1000002
    }
  ],
  "file/base": {
    "record": {
      "extension": ".pdf",
      "hash": "615fdc8a3ee8ef23e27500d86bd982eb2b76ec4920b3c81f284b1bf12464ddb6",
      "media_type": "application/pdf",
      "media_type_prefix": "application",
      "name": "2601.05776v1.pdf",
      "size": 367245
    },
    "source": {
      "execution_id": "1373bf24-562d-47d1-af8e-53aff64baeb2",
      "id": "file/base",
      "type": "Model",
      "version": "1.0.0"
    }
  },
  "file/pdf": {
    "private": false,
    "record": {
      "keywords": [],
      "page_count": 15,
      "producer": "pikepdf 8.15.1",
      "title": "One Script Instead of Hundreds? On Pretraining Romanized Encoder Language Models",
      "version": "1.7"
    },
    "source": {
      "execution_id": "56d9c3d4-f152-4921-97cc-e555738068cf",
      "id": "dorsal/pdf",
      "type": "Model",
      "version": "1.1.0"
    }
  }
}