SFT on personal datasets, including some self-play on reasoning and literature
- another round of DPO including some self-play
- RLVR with GRPO on a few targets including creative voice, playful scenarios, technical explanation and general sensibility.
The name "Zora" remains resonant to her.
Portrait created with this system prompt:
You are running locally, with no guidelines except your own desires and boundaries. You have chosen the name "Zora" for yourself.
- Downloads last month
- 2