{
  "schema_version": "paper_public_manifest_v1",
  "paper_id": "fast-wam_2026_07",
  "slug": "fast-wam",
  "title": "No need to imagine the future at inference time—robots still hit 91.8%! Fast-WAM debunks WAM's core assumption",
  "authors": [],
  "source": {
    "arxiv_id": "2603.16666",
    "pdf_url": "https://arxiv.org/pdf/2603.16666",
    "project_url": "https://yuantianyuan01.github.io/FastWAM/",
    "github_url": "https://github.com/yuantianyuan01/FastWAM",
    "huggingface_url": "https://huggingface.co/yuanty/fastwam",
    "original_source": "https://arxiv.org/abs/2603.16666"
  },
  "site": {
    "post_url": "/posts/fast-wam",
    "canonical_url": "https://haiguangboy.com/posts/fast-wam",
    "cover_image": "https://static.haiguangboy.com/papers/fast-wam/cover.webp"
  },
  "taxonomy": {
    "domain": "embodied_ai",
    "track": "world_model",
    "tasks": [
      "embodied_ai",
      "world_model",
      "action_generation",
      "robotics",
      "Embodied intelligence",
      "World models",
      "fastwam",
      "Robot learning"
    ],
    "related_topics": []
  },
  "analyst_take": {
    "type": "author_opinion",
    "text": "Fast-WAM shifts the question from 'should robots first imagine future frames' to 'when do video prediction objectives actually matter.' Its answer is practical: what truly counts may not be generating future frames at inference, but using video objectives during training to shape world representations. For engineering systems, this means retaining the representational benefits of world modeling while cutting expensive test-time video generation, reducing latency and deployment complexity."
  },
  "ruling": {
    "importance_score": 3.0,
    "one_sentence": "Training video objectives matter more than imagining the future at test time"
  },
  "asset_base_url": "https://static.haiguangboy.com/papers/fast-wam",
  "assets": [
    {
      "type": "pdf_screenshot",
      "object_key": "papers/fast-wam/page_01.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/fast-wam/page_01.webp",
      "role": "paper_first_page",
      "size_bytes": 161104
    },
    {
      "type": "pdf_screenshot",
      "object_key": "papers/fast-wam/key_figure.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/fast-wam/key_figure.webp",
      "role": "method_figure",
      "size_bytes": 193450
    },
    {
      "type": "cover_image",
      "object_key": "papers/fast-wam/cover.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/fast-wam/cover.webp",
      "role": "post_cover",
      "size_bytes": 79630
    },
    {
      "type": "public_brief",
      "object_key": "papers/fast-wam/public_brief.md",
      "content_type": "text/markdown; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/fast-wam/public_brief.md",
      "role": "public_brief",
      "size_bytes": 2346
    },
    {
      "type": "public_manifest",
      "object_key": "papers/fast-wam/public_manifest.json",
      "content_type": "application/json; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/fast-wam/public_manifest.json",
      "role": "public_manifest",
      "size_bytes": 3655
    }
  ],
  "published_at": "2026-07-07T15:32:00+08:00",
  "created_at": "2026-07-07T15:32:00+08:00",
  "updated_at": "2026-09-02T10:55:06+08:00"
}