medismera/Qwen3.8-cyber-RedTeam-Surgical-Abliterated
Partial fine-tune vs medismera/Qwen3.8-27B-Surgical-Abliterated. Partial change: 79% of the weights changed (attention, embeddings, lm_head, mlp), in 64 of 64 layers.
Tensors changed: Norms 0/319, Attention 128/764, MLP 384/496, lm_head 1/1, Embeddings 1/4