data/manifest.json

{
  "schema_version": "spavobench-v1",
  "generated_at": "2026-08-11T02:33:24Z",
  "source_repository": "Xinyun-Liu/WorldModelBench",
  "source_video_policy": "Complete source MP4s remain in WorldModelBench. This release stores stable paths and URLs without duplicating video files.",
  "timestamp_convention": "One or two digits: seconds. Three or four digits: leading digits are minutes and final two digits are seconds. Decimal timestamps are seconds.",
  "record_counts": {
    "all_reviewed": 276,
    "release_keep": 276,
    "source_videos": 248,
    "numbered_source_rows_split": 22,
    "frame_failures": 29
  },
  "by_decision": {
    "keep": 276
  },
  "by_data_source": {
    "Ego4D": 26,
    "Honda Advice Dataset (HAD)": 8,
    "PAI-Bench": 41,
    "Panda-70M": 57,
    "Physics-IQ": 22,
    "Sekai Walking": 21,
    "SpatialVID": 12,
    "SpatialVID-HQ": 9,
    "VGenST-Bench": 49,
    "WISA-80K": 31
  },
  "incremental_update": {
    "review_file": "worldmodelbench-review-2026-08-10.json",
    "source_keep_rows_added": 73,
    "image_caption_records_added": 59,
    "skipped_or_failed_records": 29
  },
  "prompt_schema": {
    "schema_version": "spavobench-prompt-schema-v1",
    "fields": [
      "perspective",
      "scene_context",
      "motion_subject",
      "camera_constraint",
      "intervention"
    ],
    "model_prompt": "A fluent public caption composed from perspective, camera_constraint, and intervention.",
    "source": "review-approved structured prompt draft"
  }
}