Null-Guard/Qwen3-0.6B-Uncensored
Abliterated vs Qwen/Qwen3-0.6B. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 28 of 28 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Norms 0/113, MLP 28/84, Attention 28/112
1 derivatives checked
- Null-Guard/Qwen3-0.6B-Uncensored-GGUF: Same weights, converted