Datasets:

Infi-MM
/

InfiMM-WebMath-40B

Tasks:

Image-Text-to-Text

Modalities:

Formats:

Languages:

Size:

ArXiv:

Tags:

mathematical-reasoning

Libraries:

License:

Dataset card Data Studio Files Files and versions

xiaotianhan commited on Sep 19, 2024

Commit

375045e

·

verified ·

1 Parent(s): 7873de8

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -21,7 +21,7 @@ size_categories:
 The **InfiMM-WebMath-40B** dataset includes:
 - **24 million** web documents.
-- **14 million** unique images.
 - **40 billion** text tokens.
 These documents were sourced from **Common Crawl** data snapshots (2019–2023), filtered to focus on high-quality mathematical and scientific content in both English and Chinese.

 The **InfiMM-WebMath-40B** dataset includes:
 - **24 million** web documents.
+- **85 million** image URLs.
 - **40 billion** text tokens.
 These documents were sourced from **Common Crawl** data snapshots (2019–2023), filtered to focus on high-quality mathematical and scientific content in both English and Chinese.