shunzou05/TerminalHorizon-27B

Full fine-tune vs Qwen/Qwen3.5-27B. Fine-tune: 94% of the weights changed, across 64 of 64 layers.

Tensors changed: lm_head 1/1, Embeddings 0/4, MLP 192/304, Attention 371/556, Norms 136/319

2 derivatives checked