pavelslab-nyu/pretrain_v1_54B

Chess Pre-to-Post — Pretraining Corpus v1 (54B)

Raw tokenized pretraining data for chess models pretraining, stored as sharded NumPy arrays (shard_XXXX/raw.NNNN.npy).

Format

• Sharded raw token arrays: shard_0000/ … shard_0005/, each containing raw.NNNN.npy files of tokenized text.
• Load shards with numpy.load(...) and concatenate as needed for your data loader.

Citation

If you use this dataset, please cite it:

@article{pre2post-chess,
  title   = {Understanding Reasoning from Pretraining to Post-Training},
  author  = {Shen, Jingyan and Li, Ang and Rahman, Salman and Sun, Yifan and
             Goldblum, Micah and Telgarsky, Matus and Izmailov, Pavel},
  journal = {arXiv preprint arXiv:2607.16097},
  year    = {2026},
  url     = {https://arxiv.org/pdf/2607.16097}
}