Training data and RL code

[EHxuban11]

Hello, thank you for this brilliant work.

I was wondering if you plan to release the training data and the code for RL.

[ohhyeeaah]

Thank you!

We are considering releasing more training details, data recipes, and RL code in the future, but we don’t have a fixed timeline yet.

What we can say for now is that the data sources we used are open-source / public datasets, with filtering, synthesis, verification, and decontamination on top. Before releasing anything, we still need to clean up the pipeline, check licenses, and make sure the released version is actually useful and reproducible.

[EHxuban11]

Hello @ohhyeeaah , thank you very much for your response.

Sure, this things are tricky. Looking forward to it!