huihui-ai/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated
Abliterated vs empero-ai/Qwythos-9B-Claude-Mythos-5-1M. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 22 of 32 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 0/215, Attention 22/332, MLP 22/208
6 derivatives checked
- mradermacher/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated-i1-GGUF: Quantization
- Abiray/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated-GGUF: Quantization
- mradermacher/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated-GGUF: Same weights, converted
- maci0/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated-NVFP4: Quantization
- LeaderboardModel1/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated-AutoRound-W4A16-RTN: Quantization
- keXjos/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated-mlx-8Bit: Quantization
3 other repos hold byte-for-byte identical weight files.