KridgeDookie/Qwen3.6-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS
Abliterated vs Qwen/Qwen3.6-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 63 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 63/304, Attention 63/556, lm_head 0/1, Norms 0/319