2601.08842v1.pdf

{
  "dorsal/arxiv": [
    {
      "date_modified": "2026-02-17T05:53:20.002000Z",
      "hash": "527a32f5142726a09091f605440178dec8c7895b66f2ccd13583d88f0c23897b",
      "id": "c52f3066-a7ef-44fb-83b7-afd926ba257b",
      "source": {
        "execution_id": "6429ebf8-cf28-468b-b4eb-ef1f59506b6a",
        "id": "arXiv Dataset",
        "type": "Model",
        "variant": "snapshot-2026-01-17",
        "version": "0.1.0"
      },
      "user_id": 1000002
    }
  ],
  "file/base": {
    "record": {
      "extension": ".pdf",
      "hash": "527a32f5142726a09091f605440178dec8c7895b66f2ccd13583d88f0c23897b",
      "media_type": "application/pdf",
      "media_type_prefix": "application",
      "name": "2601.08842v1.pdf",
      "size": 399025
    },
    "source": {
      "execution_id": "6500aaed-7b95-4a47-9082-717dc9c16c56",
      "id": "file/base",
      "type": "Model",
      "version": "1.0.0"
    }
  },
  "file/pdf": {
    "private": false,
    "record": {
      "keywords": [],
      "page_count": 6,
      "producer": "pikepdf 8.15.1",
      "title": "Resisting Correction: How RLHF Makes Language Models Ignore External Safety Signals in Natural Conversation",
      "version": "1.7"
    },
    "source": {
      "execution_id": "1d066cf4-ae18-404b-9df9-f94ff2d22583",
      "id": "dorsal/pdf",
      "type": "Model",
      "version": "1.1.0"
    }
  }
}