LLaVA-1.5-665K: 665,000 Multimodal Instruction-Following Pairs

665,000 multimodal instruction-following pairs consisting of images and text sequences, compiled by kaiyuyue and updated in 2025. This collection consolidates the LLaVA-1.5-665K mixture into a single repository, providing raw images in WebDataset format alongside instruction JSONs.

Use Cases

• Training Vision-Language Models (VLMs) using the image and text instruction pairs
• Zero-shot vision encoder grafting as demonstrated in the associated research paper
• Fine-tuning models for visual question answering using the mix665k JSON sequences

Strengths

• 665,000 multimodal instruction pairs
• Consolidated format with images and text in one repository
• CC BY 4.0 permissive license

Limitations

• Large storage requirements due to 665,000 images in WebDataset format
• Potential label noise inherent in large-scale instruction-following mixtures

Provenance

Source

kaiyuyue / LLaVA-1.5 research

Freshness

Last updated August 2025.