Commit History

  1. Update README with detailed data pipeline and reproduction steps

    Verified signature
    2d3e79d
  2. Add model card README

    Verified signature
    d471665
  3. Add eval.py

    Verified signature
    48d87be
  4. Add train_rs_sft.py

    Verified signature
    10c8d20
  5. Add rs_sample.py

    Verified signature
    1dcce72
  6. Add train_sft.py

    Verified signature
    1c4102a
  7. SFT + Rejection Sampling SFT (5x teacher replay). GSM8K avg ~46.6%, best 48.9%

    Verified signature
    8e96ac1
  8. initial commit

    Verified signature
    4f6c780