Indexnusrefather/Caelistus-Gladius-Excalibur-RP-2.6B-Thinking-v0.1

Partial fine-tune vs LiquidAI/LFM2.5-2.6B. Partial change: 22% of the weights changed (attention, mlp, other), in 26 of 30 layers.

Tensors changed: Embeddings 0/2, other 18/66, MLP 20/90, Norms 0/76, Attention 30/32

2 derivatives checked