Commit History

  1. Flag as preliminary: retraining on cleaned socrateach-sft (incorrect-math dialogues removed)

    Verified signature
    f4d55ae
  2. Pin training data to socrateach-sft v1 (math errata note)

    Verified signature
    2178a2e
  3. Impl-3 7B LoRA adapters: vanilla SFT control + 6 KL-reweighted variants

    Verified signature
    ff6cafa
  4. initial commit

    Verified signature
    d020623