OmniScience: 1M-10M Multi-modal Pairs for Scientific Image Understanding

OmniScience provides between 1 million and 10 million multi-modal records for scientific image understanding, released by UniParser in January 2026. The data pairs scientific imagery with text to support image-to-text tasks, following a collection phase completed in September 2025.

Use Cases

• Training image-to-text models to generate captions for scientific figures
• Fine-tuning multi-modal LLMs for scientific visual reasoning
• Benchmarking zero-shot performance on scientific image understanding tasks

Strengths

• Scale of 1M to 10M records
• Ranked Top 1 in Hugging Face Image Captioning field upon release
• Provided in optimized-parquet format for efficient data loading

Limitations

• CC-BY-NC-SA 4.0 license prohibits commercial use
• Scientific domain focus may limit generalizability to natural image datasets

Provenance

Source

UniParser (associated with Arxiv 2602.13758 and 2512.15098)

Time Range

Completed 2025-09-25

Freshness

Last updated March 2026; data collection was finalized in September 2025.