huihui-ai/Huihui-Qwen3-4B-abliterated-v2
Abliterated vs Qwen/Qwen3-4B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 36 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 36/108, Attention 36/144
4 derivatives checked
- mradermacher/Huihui-Qwen3-4B-abliterated-v2-GGUF: Same weights, converted
- mradermacher/Huihui-Qwen3-4B-abliterated-v2-i1-GGUF: Quantization
- tensorblock/huihui-ai_Huihui-Qwen3-4B-abliterated-v2-GGUF: Quantization
- Sonorix/gpt-aurora-v1-gguf: Quantization
1 other repos hold byte-for-byte identical weight files.