2601.19208v1.pdf

{
  "file/base": {
    "record": {
      "extension": ".pdf",
      "hash": "23ac80fabdc6aeab76e1c7c3470374d64c6a618e610ccf981cd7f142e89251e1",
      "media_type": "application/pdf",
      "media_type_prefix": "application",
      "name": "2601.19208v1.pdf",
      "size": 1142089
    },
    "source": {
      "execution_id": "eb5777a7-efd0-48be-b655-302c246a696e",
      "id": "file/base",
      "type": "Model",
      "version": "1.0.0"
    }
  },
  "file/pdf": {
    "private": false,
    "record": {
      "keywords": [],
      "page_count": 36,
      "producer": "pikepdf 8.15.1",
      "title": "How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability",
      "version": "1.7"
    },
    "source": {
      "execution_id": "e66a47dc-851b-4b51-a61a-80cb11c483b1",
      "id": "dorsal/pdf",
      "type": "Model",
      "version": "1.1.0"
    }
  }
}