amkkk/Qwen3.5-4B-abliterated-aggressive
Abliterated vs Qwen/Qwen3.5-4B. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj, out_proj) changed, in 32 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/4, MLP 32/196, Attention 32/320, Norms 0/203
2 derivatives checked
- mradermacher/Qwen3.5-4B-abliterated-aggressive-i1-GGUF: Quantization
- mradermacher/Qwen3.5-4B-abliterated-aggressive-GGUF: Same weights, converted