{
  "schema_version": 1,
  "annotation_revision": 2,
  "annotation_status": "user_confirmed_release",
  "source_dataset": {
    "repo_id": "wzlllong/value-model-benchmark-training-data",
    "subtree": "data/simulation/one-shot",
    "data_revision": "43ae4845fe0461c66aa59a0ee441185d99fb05f8",
    "source_subtree_tree_id": "837e271aac36ea3eea72599ab5bca86db64bb025"
  },
  "dataset": {
    "codebase_version": "v3.0",
    "fps": 25,
    "total_episodes": 15,
    "total_frames": 8080,
    "total_videos": 45
  },
  "paper": {
    "url": "https://arxiv.org/abs/2610.26663",
    "revision": "2610.26663v1",
    "section": "C.2 Task Descriptions and Subtask Decomposition"
  },
  "segment_count": 58,
  "paper_subtask_count": 58,
  "annotation_target": "Active semantic subtask for each training-demonstration frame; not success, reward, or task-completion probability.",
  "release_scope": "Supplemental annotations for simulation One-Shot training demonstrations, separate from benchmark test-set labels described in paper Section 3.4.",
  "coverage_policy": "Every frame receives one printed paper subtask. Initial waiting belongs to the first subtask; terminal tail and retraction belong to the final subtask through episode end.",
  "interval_convention": "[start_frame, end_frame); inclusive_end_frame = end_frame - 1",
  "join_keys": [
    "episode_index",
    "frame_index"
  ],
  "global_index_check": "global_index equals the LeRobot data index column",
  "timestamp_convention": "Nominal timestamp_seconds = frame_index / 25; associate frames by integer keys, not exact float timestamp equality.",
  "boundary_band": "Typically 13 frames (0.52 seconds); retained segment-specific widths are suggested adjustment bands, not measured confidence bounds or guarantees of frame-perfect precision.",
  "boundary_uncertain_policy": "For each segment, mark its first t frames after an internal start and last t frames before an internal end; episode outer boundaries excluded. Keep all frame labels regardless of mask.",
  "video_loading": {
    "recommended_lerobot": "0.4.2",
    "recommended_pyav": "15.1.0",
    "default_tolerance_seconds": 0.0001,
    "semantic_band_changes_loader_tolerance": false
  },
  "source_training_payload_modified": false,
  "printed_label_policy": "Use the paper's printed S definitions, including Press By Number's S1 first button, S2 blue, S3 second button, S4 blue again; preserve the accepted order without rewriting the paper instruction.",
  "task_aliases": {
    "sweep_objects": "sweep_blocks"
  },
  "episodes": [
    {
      "episode_index": 0,
      "task_index": 0,
      "task_key": "arrange_largest_number",
      "source_task_key": "arrange_largest_number",
      "length": 579,
      "dataset_from_index": 0,
      "dataset_to_index": 579
    },
    {
      "episode_index": 1,
      "task_index": 1,
      "task_key": "fill_pen_holder",
      "source_task_key": "fill_pen_holder",
      "length": 934,
      "dataset_from_index": 579,
      "dataset_to_index": 1513
    },
    {
      "episode_index": 2,
      "task_index": 2,
      "task_key": "fold_clothes",
      "source_task_key": "fold_clothes",
      "length": 301,
      "dataset_from_index": 1513,
      "dataset_to_index": 1814
    },
    {
      "episode_index": 3,
      "task_index": 3,
      "task_key": "hang_mugs",
      "source_task_key": "hang_mugs",
      "length": 513,
      "dataset_from_index": 1814,
      "dataset_to_index": 2327
    },
    {
      "episode_index": 4,
      "task_index": 4,
      "task_key": "insert_tubes",
      "source_task_key": "insert_tubes",
      "length": 320,
      "dataset_from_index": 2327,
      "dataset_to_index": 2647
    },
    {
      "episode_index": 5,
      "task_index": 5,
      "task_key": "make_toast",
      "source_task_key": "make_toast",
      "length": 584,
      "dataset_from_index": 2647,
      "dataset_to_index": 3231
    },
    {
      "episode_index": 6,
      "task_index": 6,
      "task_key": "organize_table",
      "source_task_key": "organize_table",
      "length": 839,
      "dataset_from_index": 3231,
      "dataset_to_index": 4070
    },
    {
      "episode_index": 7,
      "task_index": 7,
      "task_key": "pack_objects_into_box",
      "source_task_key": "pack_objects_into_box",
      "length": 414,
      "dataset_from_index": 4070,
      "dataset_to_index": 4484
    },
    {
      "episode_index": 8,
      "task_index": 8,
      "task_key": "play_stacking_toy",
      "source_task_key": "play_stacking_toy",
      "length": 888,
      "dataset_from_index": 4484,
      "dataset_to_index": 5372
    },
    {
      "episode_index": 9,
      "task_index": 9,
      "task_key": "press_by_number",
      "source_task_key": "press_by_number",
      "length": 378,
      "dataset_from_index": 5372,
      "dataset_to_index": 5750
    },
    {
      "episode_index": 10,
      "task_index": 10,
      "task_key": "put_bottles_into_dustbin",
      "source_task_key": "put_bottles_into_dustbin",
      "length": 363,
      "dataset_from_index": 5750,
      "dataset_to_index": 6113
    },
    {
      "episode_index": 11,
      "task_index": 11,
      "task_key": "stack_blocks",
      "source_task_key": "stack_blocks",
      "length": 268,
      "dataset_from_index": 6113,
      "dataset_to_index": 6381
    },
    {
      "episode_index": 12,
      "task_index": 12,
      "task_key": "store_laptop_and_headphones",
      "source_task_key": "store_laptop_and_headphones",
      "length": 551,
      "dataset_from_index": 6381,
      "dataset_to_index": 6932
    },
    {
      "episode_index": 13,
      "task_index": 13,
      "task_key": "swap_blocks",
      "source_task_key": "swap_blocks",
      "length": 445,
      "dataset_from_index": 6932,
      "dataset_to_index": 7377
    },
    {
      "episode_index": 14,
      "task_index": 14,
      "task_key": "sweep_blocks",
      "source_task_key": "sweep_objects",
      "length": 703,
      "dataset_from_index": 7377,
      "dataset_to_index": 8080
    }
  ]
}