YOYO-AI/Qwen2.5-14B-YOYO-stock

Full fine-tune vs Gen-Verse/ReasonFlux-F1-14B. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Norms 97/97, MLP 144/144, Attention 336/336

1 derivatives checked