Several questions on OpenDoc-0.1B

[alephpi]

Hi, thank you for open sourcing this valuable work, it's surprising to have such tiny model that beats the VLM giants.

I notice that the latest OpenDoc-0.1B is available on HF, and I also find an empty dataset UniRec-40M in your HF space. Would you plan to open source it too?

Besides, does OpenDoc-0.1B follow the same training recipe as UniRec-0.1B? Do you plan to release the training code too?

Also, you claim OpenDoc-0.1B performs over 90% on OmniDocBench-1.5, would you also provide detailed scores as they did?

Sincerely,
alephpi