windowsxp811203/Qwen3.8-27B-Abliterated
Abliterated vs Qwen/Qwen3.8-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj, out_proj) changed, in 64 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/325, Attention 65/560, MLP 65/307, Embeddings 1/5, lm_head 0/1, other 0/1
9 derivatives checked
- windowsxp811203/Qwen3.8-27B-Abliterated-GGUF: Quantization
- finex666/Qwen3.8-27B-Abliterated-IQ4-MIX-MTP-GGUF: Quantization
- windowsxp811203/Qwen3.8-27B-Abliterated-MLX-bf16-mtp: Partial fine-tune
- windowsxp811203/Qwen3.8-27B-Abliterated-MLX-oQ4e-mtp: Quantization
- windowsxp811203/Qwen3.8-27B-Abliterated-MLX-oQ8e-mtp: Quantization
- windowsxp811203/Qwen3.8-27B-Abliterated-MTP-bf16: Partial fine-tune (fewer layers)
- windowsxp811203/Qwen3.8-27B-Abliterated-MLX-oQ6e-mtp: Quantization
- windowsxp811203/Qwen3.8-27B-Abliterated-NVFP4: Quantization
- windowsxp811203/Qwen3.8-27B-Abliterated-NVFP4-v2: Quantization