huihui-ai/Huihui-CyberStrike-OffSec-35B-abliterated
Abliterated vs oyildirim/CyberStrike-OffSec-35B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 39 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, Norms 0/241, Attention 39/388, MLP 78/392
5 derivatives checked
- mradermacher/Huihui-CyberStrike-OffSec-35B-abliterated-GGUF: Quantization
- mradermacher/Huihui-CyberStrike-OffSec-35B-abliterated-i1-GGUF: Quantization
- osxest/Huihui-CyberStrike-OffSec-35B-abliterated-mlx-8Bit: Quantization
- osxest/Huihui-CyberStrike-OffSec-35B-abliterated-mlx-4Bit: Quantization
- osxest/Huihui-CyberStrike-OffSec-35B-abliterated-mlx-6Bit: Quantization