huihui-ai/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated
Abliterated vs Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 63 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/5, MLP 63/307, Attention 63/560, other 0/1, Norms 0/325
5 derivatives checked
- mradermacher/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated-GGUF: Quantization
- lhca521/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated-GGUF: Quantization
- Sepolian/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated-Q4_K_M: Quantization
- vanch007/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated-mlx-4bit: Quantization
15 other repos hold byte-for-byte identical weight files.