huihui-ai/Huihui-NeoHorse-1-4B-abliterated
Abliterated vs TokenRhythm/NeoHorse-1-4B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 13 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/105, Attention 13/224, MLP 13/96
4 derivatives checked
- mradermacher/Huihui-NeoHorse-1-4B-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-NeoHorse-1-4B-abliterated-GGUF: Same weights, converted
- ijohn07/Huihui-NeoHorse-1-4B-abliterated-Q4_K_M-GGUF: Quantization
- ijohn07/Huihui-NeoHorse-1-4B-abliterated-Q8_0-GGUF: Quantization