2601.21666v1.pdf

{
  "file/base": {
    "record": {
      "extension": ".pdf",
      "hash": "a2d798b5166095b32ec36ccfd73f962744a6740efeea4080e89f5fb494fd0337",
      "media_type": "application/pdf",
      "media_type_prefix": "application",
      "name": "2601.21666v1.pdf",
      "size": 7064399
    },
    "source": {
      "execution_id": "8ed541bd-b81d-4120-b4ec-7326eedca7aa",
      "id": "file/base",
      "type": "Model",
      "version": "1.0.0"
    }
  },
  "file/pdf": {
    "private": false,
    "record": {
      "keywords": [],
      "page_count": 47,
      "producer": "pikepdf 8.15.1",
      "title": "SONIC-O1: A Real-World Benchmark for Evaluating Multimodal Large Language Models on Audio-Video Understanding",
      "version": "1.7"
    },
    "source": {
      "execution_id": "48449719-d9e1-4019-ac05-6775ef38f588",
      "id": "dorsal/pdf",
      "type": "Model",
      "version": "1.1.0"
    }
  }
}