2601.11652v1.pdf

{
  "file/base": {
    "record": {
      "extension": ".pdf",
      "hash": "8f8d80b3b62dccb034f016bd2cbb168f88726bec9e0656a153c73679521c2f6b",
      "media_type": "application/pdf",
      "media_type_prefix": "application",
      "name": "2601.11652v1.pdf",
      "size": 2942436
    },
    "source": {
      "execution_id": "c5edb489-1007-49f6-bb96-3c82d7ddac10",
      "id": "file/base",
      "type": "Model",
      "version": "1.0.0"
    }
  },
  "file/pdf": {
    "private": false,
    "record": {
      "creation_date": "2026-01-21T01:08:03Z",
      "keywords": [
        "Speculative Decoding",
        "Large Language Models",
        "Edge Computing",
        "Distributed Inference",
        "Token Verification",
        "Resource-Aware Serving"
      ],
      "modified_date": "2026-01-21T01:08:03Z",
      "page_count": 29,
      "producer": "pikepdf 8.15.1",
      "subject": "-  Computing methodologies  -\u003e  Distributed computing methodologies.Distributed artificial intelligence.Natural language processing.-  Computer systems organization  -\u003e  Distributed architectures.-  Networks  -\u003e  Cloud computing.-  General and reference  -\u003e  General conference proceedings.",
      "title": "WISP: Waste- and Interference-Suppressed Distributed Speculative LLM Serving at the Edge via Dynamic Drafting and SLO-Aware Batching",
      "version": "1.7"
    },
    "source": {
      "execution_id": "cb86ea70-5d90-4f3c-910d-e91472b49ed4",
      "id": "dorsal/pdf",
      "type": "Model",
      "version": "1.1.0"
    }
  }
}