{
  "schema_version": "paper_public_manifest_v1",
  "paper_id": "orca_2026_07",
  "slug": "orca",
  "title": "π0.5 trembles in place after failing to grab a spoon, but Orca's physics intuition, learned from watching videos, goes further",
  "authors": [],
  "source": {
    "arxiv_id": "2606.30534",
    "pdf_url": "https://arxiv.org/pdf/2606.30534",
    "project_url": "https://orca-wm.github.io/",
    "github_url": "",
    "original_source": "https://arxiv.org/abs/2606.30534"
  },
  "site": {
    "post_url": "/posts/orca",
    "canonical_url": "https://haiguangboy.com/posts/orca",
    "cover_image": "https://static.haiguangboy.com/papers/orca/cover.webp"
  },
  "taxonomy": {
    "domain": "embodied_ai",
    "track": "world_model",
    "tasks": [
      "embodied_ai",
      "world_model",
      "vla",
      "action_generation",
      "robotics",
      "latent_state",
      "state_prediction",
      "Embodied Intelligence",
      "World Models"
    ],
    "related_topics": [
      {
        "paper_id": "an_open_foundation_model_towards_2026_07",
        "title": "An_Open_Foundation_Model_Towards",
        "url": "https://haiguangboy.com/posts/an_open_foundation_model_towards",
        "relation": "same_track",
        "summary": "Action generation comparison: enabling Qwen3.5 to break through from 0% success rate and match π0.5 pretrained on large-scale robot data validates that joint co-training on heterogeneous human-robot data is a structurally suboptimal approach",
        "strength": "strong"
      },
      {
        "paper_id": "fast-wam_2026_07",
        "title": "No need to imagine the future at inference time, robot still hits 91.8%! Fast-WAM debunks WAM's core assumption",
        "url": "https://haiguangboy.com/posts/fast-wam",
        "relation": "same_track",
        "summary": "Both show that the value of world models lies not in generating the future at inference time, but in shaping transferable world representations during training.",
        "strength": "strong"
      }
    ]
  },
  "ruling": {
    "importance_score": 3.0,
    "one_sentence": "The key to a world model is a readable state"
  },
  "asset_base_url": "https://static.haiguangboy.com/papers/orca",
  "assets": [
    {
      "type": "pdf_screenshot",
      "object_key": "papers/orca/page_01.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/orca/page_01.webp",
      "role": "paper_first_page",
      "size_bytes": 167482
    },
    {
      "type": "pdf_screenshot",
      "object_key": "papers/orca/key_figure.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/orca/key_figure.webp",
      "role": "method_figure",
      "size_bytes": 173794
    },
    {
      "type": "cover_image",
      "object_key": "papers/orca/cover.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/orca/cover.webp",
      "role": "post_cover",
      "size_bytes": 76828
    },
    {
      "type": "public_brief",
      "object_key": "papers/orca/public_brief.md",
      "content_type": "text/markdown; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/orca/public_brief.md",
      "role": "public_brief",
      "size_bytes": 3759
    },
    {
      "type": "public_manifest",
      "object_key": "papers/orca/public_manifest.json",
      "content_type": "application/json; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/orca/public_manifest.json",
      "role": "public_manifest",
      "size_bytes": 3875
    }
  ],
  "created_at": "2026-07-06T23:19:24+08:00",
  "updated_at": "2026-09-02T10:55:06+08:00",
  "published_at": "2026-07-06T23:19:24+08:00",
  "analyst_take": {
    "type": "author_opinion",
    "text": "The finding that Orca enables Qwen3.5 to break through from 0% success rate and match π0.5 pretrained on large-scale robot data is judged in cross-library relational metabolism to strongly corroborate several other papers: Fast-WAM achieves near-SOTA without embodied pretraining, Ψ0 surpasses a 10x larger data baseline by over 40% with only a small amount of data, and General Flow achieves zero-shot high-success-rate transfer via human video pretraining. Several papers with completely different technical routes are converging on the same direction: what is truly scarce is not the quantity of robot action data, but whether one can first learn a high-quality world representation."
  }
}