Datasets:

mvp-lab
/

LLaVA-OneVision-2-Data

license: apache-2.0
task_categories:
  - video-text-to-text
  - visual-question-answering
  - image-text-to-text
language:
  - en
tags:
  - llava
  - multimodal
  - video
  - spatial-reasoning
size_categories:
  - 10M<n<100M
configs:
  - config_name: viewer_caption_30s
    data_files:
      - split: preview
        path: viewer/caption_30s.parquet
  - config_name: viewer_caption_60s
    data_files:
      - split: preview
        path: viewer/caption_60s.parquet
  - config_name: viewer_spatial
    data_files:
      - split: preview
        path: viewer/spatial.parquet

LLaVA-OneVision-2-Data

Training data for the LLaVA-OneVision-2 multimodal model family, covering large-scale video and spatial reasoning corpora used in mid-training.

Dataset Composition

Subset	Format	Description
`mid_training_video/60s_rest/`	WebDataset (`.tar`)	10,809 shards of ~60s video clips
`mid_training_video/caption_v0/split_30s.jsonl`	JSONL	Captions for 30-second video clips
`mid_training_video/caption_v0/split_60s.jsonl`	JSONL	Captions for 60-second video clips
`spatial/`	WebDataset (`.tar`)	84 shards of spatial reasoning data (refcoco, visual genome, pointing, 3D, etc.)

Preview Configs

The three viewer_* configs above expose small Parquet samples so the Hugging Face Dataset Viewer can render the data directly in the browser:

viewer_caption_30s — 5 caption samples from 30-second clips
viewer_caption_60s — 5 caption samples from 60-second clips
viewer_spatial — 10 spatial-reasoning samples with embedded thumbnail images, mixed across tasks (refcoco, visual genome, pointing, ca1m, osd, crosspoint, erqa, roborefer)

These previews are intended for schema inspection only. For training, use the full mid_training_video/ and spatial/ shards.

Loading Examples

Stream the full spatial WebDataset

from datasets import load_dataset

ds = load_dataset(
    "mvp-lab/LLaVA-OneVision-2-Data",
    data_files="spatial/train_*.tar",
    split="train",
    streaming=True,
)
for sample in ds:
    print(sample)
    break

Load a viewer preview

from datasets import load_dataset

ds = load_dataset("mvp-lab/LLaVA-OneVision-2-Data", "viewer_spatial", split="preview")
print(ds[0])

Schema

Caption (`viewer_caption_30s`, `viewer_caption_60s`)

images_source: List[str] — relative paths to source video files
messages: List[{role, content}] — chat-format caption prompt and assistant response
preview: str — flattened text preview

Spatial (`viewer_spatial`)

id: str — sample id
task: str — source task name (e.g. refcoco, visual_genome, pointing)
image: Image — first referenced image, embedded as JPEG bytes (downscaled to max edge 1024)
image_paths: List[str] — original relative image paths in the full dataset
messages: List[{role, content}] — chat-format Q/A
preview: str — flattened text preview

License

Apache 2.0