roslein/Qwen3-32B-abliterated
Abliterated vs Qwen/Qwen3-32B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 60 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/257, MLP 60/192, Attention 60/256, lm_head 0/1
3 derivatives checked
- mradermacher/Qwen3-32B-abliterated-GGUF: Quantization
- mradermacher/Qwen3-32B-abliterated-i1-GGUF: Quantization
- Thaurock/Qwen3-32B-abliterated-GGUF: Same weights, converted