LLaVA-CC3M-Pretrain-595K: Thai-Translated Multimodal Pretraining Data

A Thai translation of the LLaVA-CC3M-Pretrain-595K dataset, originally created by Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. This dataset is intended for pre-training large multimodal models with Thai language capabilities and was uploaded to Hugging Face by user 'worapob' on March 9, —2025.

Use Cases

• Pretraining multimodal models based on the Thai-translated image-text pairs.
• Fine-tuning visual question answering models for the Thai language.
• Benchmarking Thai language understanding in multimodal AI systems.

Strengths

• Dataset is a direct translation of the established LLaVA-CC3M-Pretrain-595K dataset.
• Explicitly designed for a specific, high-demand task: Thai-language multimodal model pretraining.
• Last updated on Hugging Face on 2025-03-09, indicating recent availability.

Limitations

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

Provenance

Source

Hugging Face user 'worapob', translating the original dataset by Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee.

Collection Method

Translation of the original LLaVA-CC3M-Pretrain-595K dataset.

Time Range

null

Freshness

Last updated 2025-03-09 09:01:46; freshness should be verified.

Geography

null