huihui-ai/Huihui-Qwen3.5-9B-abliterated
Abliterated vs Qwen/Qwen3.5-9B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 31 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/5, MLP 31/211, Attention 31/336, other 0/1, Norms 0/221
27 derivatives checked
- mradermacher/Huihui-Qwen3.5-9B-abliterated-GGUF: Same weights, converted
- mradermacher/Huihui-Qwen3.5-9B-abliterated-i1-GGUF: Quantization
- Abiray/Huihui-Qwen3.5-9B-abliterated-GGUF: Quantization
- RX5950XT/silicon-based-girlfriend-v2-GGUF: Full fine-tune
- nuofang/Huihui-Qwen3.5-9B-abliterated-GGUF: Not comparable yet
- td-builder/Qwen3.5-9B-abliterated-GGUF: Same weights, converted
- Kausik-A/Huihui-Qwen3.5-9B-abliterated-GGUF: Same weights, converted
- vanch007/Huihui-Qwen3.5-9B-abliterated-mlx-4bit: Quantization
- nemozxy123/Huihui-Qwen3.5-9B-abliterated-AWQ-4bit: Quantization
- nicklas373/Huihui-Qwen3.5-9B-abliterated-AWQ: Quantization
- interimlabs/InterimLabs-Huihui-Qwen3.5-9B-abliterated-GGUF-Q4_K_M: Quantization
- groxaxo/Huihui-Qwen3.5-9B-abliterated-GPTQ-Pro-4bit-g64: Quantization
- vanch007/Huihui-Qwen3.5-9B-abliterated-mlx-8bit: Quantization
- dannylr574/Huihui-Qwen3.5-9B-abliterated-Q4_K_M-GGUF: Quantization
- Rom3rk/Qwen3.5-9B-ComfyUI-GGUF: LoRA-style
- Nullmega/discord-qwen3.5-9b-gguf: LoRA-style
- divab658/Huihui-Qwen3.5-9B-abliterated-Q4_K_M-GGUF: Quantization
- interimlabs/InterimLabs-Huihui-Qwen3.5-9B-abliterated-MLX-4bit: Quantization
- Nullmega/discord-qwen3.5-9b-v2-gguf: Quantization
- anirudhsharma123/Qwen3.5-9B-abliterated-MLX-4bit: Quantization
- abcdsystems/Qwen3.5-9B-Abliterated-GDN-Hybrid-EXL3-6bpw: Same weights, converted
- Aquiles-ai/Harvey-9B: LoRA-style
- nbeerbower/Huihui-Qwen3.5-9B-abliterated-Grimoire-SFT: LoRA-style (fewer layers)
- nbeerbower/Huihui-Qwen3.5-9B-abliterated-Grimoire-SimPO: LoRA-style (fewer layers)
- nbeerbower/Huihui-Qwen3.5-9B-abliterated-Grimoire-KTO: LoRA-style (fewer layers)
- nbeerbower/Huihui-Qwen3.5-9B-abliterated-Grimoire-ORPO: LoRA-style (fewer layers)
- nbeerbower/Huihui-Qwen3.5-9B-abliterated-Grimoire-DPO: LoRA-style (fewer layers)
3 other repos hold byte-for-byte identical weight files.