{
  "schema_version": "paper_public_manifest_v1",
  "paper_id": "foveavla_2026_10",
  "slug": "foveavla",
  "title": "Without Wrist Cameras, VLA Success Rate Is Actually Higher",
  "authors": [],
  "source": {
    "arxiv_id": "2610.20064",
    "pdf_url": "https://arxiv.org/pdf/2610.20064",
    "project_url": "",
    "github_url": "",
    "huggingface_url": "",
    "original_source": "https://arxiv.org/pdf/2610.20064"
  },
  "site": {
    "post_url": "/posts/foveavla",
    "canonical_url": "https://haiguangboy.com/posts/foveavla",
    "cover_image": "https://static.haiguangboy.com/papers/foveavla/cover.webp"
  },
  "taxonomy": {
    "domain": "embodied_ai",
    "track": "vla",
    "tasks": [
      "embodied_ai",
      "vla",
      "action_generation",
      "robotics",
      "Embodied Intelligence",
      "Robotics",
      "foveavla",
      "Active Vision"
    ],
    "related_topics": [
      {
        "paper_id": "agile_wam_an_agile_tactile_world_action_model_for_contact_rich_robot_control_2026_09",
        "title": "Agile-WAM: Lightweight Visuo-Tactile Joint Prediction",
        "url": "https://haiguangboy.com/posts/agile-wam",
        "relation": "same_track",
        "summary": "48 real-robot full success runs cannot be extrapolated to industrial reliability; validates that small-sample and best-of-selection evaluations cannot be extrapolated to stable general capability",
        "strength": "strong"
      },
      {
        "paper_id": "d4rt_efficiently_reconstructing_dynamic_scenes_one_2026_06",
        "title": "D4RT_Efficiently_Reconstructing_Dynamic_Scenes_One",
        "url": "https://haiguangboy.com/posts/d4rt_efficiently_reconstructing_dynamic_scenes_one",
        "relation": "same_track",
        "summary": "Crops from original resolution while preserving the full panorama; validates that it can handle in-the-wild videos, and continuous query coordinates support decoding at arbitrary/sub-pixel resolutions",
        "strength": "strong"
      },
      {
        "paper_id": "dextouch_wm_learning_action_conditioned_tactile_world_models_from_human_touch_fo_2026_09",
        "title": "DexTouch-WM: Human Tactile Pretraining",
        "url": "https://haiguangboy.com/posts/dextouch-wm",
        "relation": "same_track",
        "summary": "Learning from human wrist-camera-free video remains future work; validates that human data is not a direct transfer of hardware-free, robot-data-free data",
        "strength": "strong"
      },
      {
        "paper_id": "google_blog_gemini_robotics_er_2_20260730_2026_07",
        "title": "Introducing Gemini Robotics ER 2",
        "url": "https://haiguangboy.com/posts/google_blog_gemini_robotics_er_2_20260730",
        "relation": "same_track",
        "summary": "Zero-shot leaderboard comparisons are not controlled comparisons; validates that the numbers lack controls: competitors are not named, baseline values are not given, and comparison charts are not verifiable",
        "strength": "strong"
      },
      {
        "paper_id": "mt_wam_reorienting_the_one_pass_predictive_representation_toward_action_generati_2026_09",
        "title": "MT-WAM: How Future Supervision Serves Action",
        "url": "https://haiguangboy.com/posts/mt-wam",
        "relation": "same_track",
        "summary": "The main results are not single-variable experiments that only remove the camera or only add cropping; validates that the causal contributions of supervision targets, capacity, and branch structure have not yet been fully disentangled",
        "strength": "strong"
      },
      {
        "paper_id": "smoothrl_online_reinforcement_learning_during_asynchronous_execution_2026_09",
        "title": "SmoothRL: Only Account for Actions Actually Executed",
        "url": "https://haiguangboy.com/posts/smoothrl_online_reinforcement_learning_during_asynchronous_execution",
        "relation": "same_track",
        "summary": "SmoothRL: Only Account for Actions Actually Executed",
        "strength": "strong"
      }
    ]
  },
  "analyst_take": {
    "type": "author_opinion",
    "text": "💡 Putting It Back into the Technical Lineage of Prediction and Control\nMT-WAM reminds us: prediction tasks being useful does not mean all prediction tokens should directly become action conditions. FoveaVLA offers another connection: prediction results change the visual details collected next time.\n\nWhat is worth testing is whether this \"predict—observe—act\" loop can sustain gains under occlusion, camera position changes, and larger workspaces, rather than merely predicting target boxes more accurately."
  },
  "ruling": {
    "importance_score": 3.0,
    "one_sentence": "Without Wrist Cameras, VLA Success Rate Is Actually Higher"
  },
  "asset_base_url": "https://static.haiguangboy.com/papers/foveavla",
  "assets": [
    {
      "type": "cover_image",
      "object_key": "papers/foveavla/cover.webp",
      "url": "https://static.haiguangboy.com/papers/foveavla/cover.webp",
      "content_type": "image/webp",
      "upload_status": "pending",
      "role": "post_cover"
    },
    {
      "type": "pdf_screenshot",
      "object_key": "papers/foveavla/page_01.webp",
      "url": "https://static.haiguangboy.com/papers/foveavla/page_01.webp",
      "content_type": "image/webp",
      "upload_status": "pending",
      "role": "paper_first_page"
    },
    {
      "type": "pdf_screenshot",
      "object_key": "papers/foveavla/key_figure.webp",
      "url": "https://static.haiguangboy.com/papers/foveavla/key_figure.webp",
      "content_type": "image/webp",
      "upload_status": "pending",
      "role": "method_figure"
    },
    {
      "type": "public_manifest",
      "object_key": "papers/foveavla/public_manifest.json",
      "url": "https://static.haiguangboy.com/papers/foveavla/public_manifest.json",
      "content_type": "application/json; charset=utf-8",
      "upload_status": "pending",
      "role": "public_manifest"
    },
    {
      "type": "public_brief",
      "object_key": "papers/foveavla/public_brief.md",
      "url": "https://static.haiguangboy.com/papers/foveavla/public_brief.md",
      "content_type": "text/markdown; charset=utf-8",
      "upload_status": "pending",
      "role": "public_brief"
    }
  ],
  "published_at": "2026-10-17T19:17:13+08:00",
  "created_at": "2026-10-17T19:17:13+08:00",
  "updated_at": "2026-10-17T19:17:13+08:00"
}