2601.10498v1.pdf

{
  "dorsal/arxiv": [
    {
      "date_modified": "2026-02-17T05:53:24.393000Z",
      "hash": "8be14da9c2daf487ac33e71617681348f55a1cfcb17029dae6076e1d9c34fe9d",
      "id": "52f7a25d-f7fe-402b-8c7d-9ccd9dcd8c2a",
      "source": {
        "execution_id": "b9ad5542-77da-462f-a767-2585ad08f9f6",
        "id": "arXiv Dataset",
        "type": "Model",
        "variant": "snapshot-2026-01-17",
        "version": "0.1.0"
      },
      "user_id": 1000002
    }
  ],
  "file/base": {
    "record": {
      "extension": ".pdf",
      "hash": "8be14da9c2daf487ac33e71617681348f55a1cfcb17029dae6076e1d9c34fe9d",
      "media_type": "application/pdf",
      "media_type_prefix": "application",
      "name": "2601.10498v1.pdf",
      "size": 589536
    },
    "source": {
      "execution_id": "a9fe8d6e-aaf3-4f3c-9645-9c95483d4f5b",
      "id": "file/base",
      "type": "Model",
      "version": "1.0.0"
    }
  },
  "file/pdf": {
    "private": false,
    "record": {
      "keywords": [],
      "page_count": 4,
      "producer": "pikepdf 8.15.1",
      "title": "Projected Microbatch Accumulation yields reference-free proximal policy updates for reinforcement learning",
      "version": "1.7"
    },
    "source": {
      "execution_id": "0362f237-c630-44a8-b870-806268c7fbb1",
      "id": "dorsal/pdf",
      "type": "Model",
      "version": "1.1.0"
    }
  }
}