null-space/Qwen3-235B-A22B-abliterated
Abliterated vs Qwen/Qwen3-235B-A22B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 73 of 94 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, MLP 539/2245, Norms 0/377, Attention 73/376, lm_head 0/1
2 derivatives checked
- mradermacher/Qwen3-235B-A22B-abliterated-i1-GGUF: Quantization
- mradermacher/Qwen3-235B-A22B-abliterated-GGUF: Quantization