{
  "absUrl": "https://arxiv.org/abs/2609.38878v1",
  "arxivId": "2609.38878",
  "arxivVersion": 1,
  "arxivVersionedId": "2609.38878v1",
  "authors": [
    {
      "affiliations": [
        "The University of Texas at Austin"
      ],
      "name": "Kyoungjun Park"
    },
    {
      "affiliations": [
        "The University of Texas at Austin"
      ],
      "name": "Yunzhe Li"
    },
    {
      "affiliations": [
        "The University of Texas at Austin"
      ],
      "name": "Lili Qiu"
    }
  ],
  "contract": "researcher-sidecars-v1",
  "id": "2609.38878v1",
  "originalTitle": "Audio Token Attention Is Predictable Before the Language Model Runs",
  "pdfUrl": "https://arxiv.org/pdf/2609.38878v1.pdf",
  "schemaVersion": 1,
  "type": "preprint"
}