Hugging Face's logo

Datasets:
ShadenA
/
MathNet

Tasks:
Question Answering
Text Generation
Image-to-Text
Modalities:
Image
Text
Formats:
parquet
Languages:
English
Portuguese
Spanish
Size:
10K - 100K
ArXiv:
Tags:
mathematics
olympiad
reasoning
competition-math
multimodal
retrieval
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
2

v1 of MathNet

#2
by FA5I - opened
Discussion
FA5I

Hi @ShadenA ,

I can see a new version was uploaded on Friday 24 April. Can I confirm that is indeed v1?

I am just reading the dataset card and it seems like it suggests the v1 dataset is yet to be updated.

Any guidance is much appreciated.
ShadenA
 Owner

Hi @FA5I ! No, I haven’t uploaded v1 yet. I was adding additional verification layers for the OCR, just in case dots.mocr makes an error. I tried a few options (gemini-2.5-flash, mathpix, GPT-4.1), but nothing has worked better than dots.mocr so far, so we’re trying to find a way to aggregate them. Hopefully, I’ll be able to share v1 very soon. Sorry for the delay.
FA5I

Thanks @ShadenA - that makes complete sense, no worries at all. I fully appreciate the difficulty of getting verification of the extracted text/latex working at scale here.

Will be following the developments with interest!

FA5I changed discussion status to closed

Sign up or log in to comment