Mobile-O Pre-Train: 9 Million Text-Image Pairs for Cross-Modal Alignment

Amshaker's dataset provides 9 million text-image pairs for the first-stage pre-training of the Mobile-O multimodal model. The data is intended to align a diffusion decoder and conditioning projector with a frozen vision-language backbone. The dataset was last updated on Hugging Face in February 2026.

Use Cases

• Pre-training vision-language models for on-device use based on the described cross-modal alignment task.
• Aligning diffusion decoders with text encoders using large-scale image-text pairs mentioned in the description.
• Training conditioning projectors for multimodal generation tasks referenced in the dataset's purpose.

Strengths

• Contains 9 million text-image pairs for large-scale training.
• Designed for a specific, documented pre-training stage (Stage 1: Cross-Modal Alignment).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Hugging Face user Amshaker.

Collection Method

Likely aggregated from multiple sources for pre-training, as suggested by the platform tags.

Time Range

null

Freshness

Last updated 2026-02-24 06:16:41; freshness should be verified.

Geography

null