philbert440/Qwen3.8-27B-Uncensored-Aggressive
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 64/560, MLP 64/307, Embeddings 1/5, lm_head 0/1, other 0/1
10 derivatives checked
- philbert440/Qwen3.8-27B-Uncensored-Aggressive-W4A16-AWQ: Quantization
- philbert440/Qwen3.8-27B-Uncensored-Cyber: Abliterated
- mradermacher/Qwen3.8-27B-Uncensored-Aggressive-GGUF: Full fine-tune
- aldenw/Qwen3.8-27B-Uncensored-Aggressive-i1-IQ4_XS-Smaller-GGUF: Quantization
- philbert440/Qwen3.8-27B-Uncensored-Aggressive-GGUF: Same weights, converted
- mradermacher/Qwen3.8-27B-Uncensored-Aggressive-i1-GGUF: Quantization
- philbert440/Qwen3.8-27B-Uncensored-Aggressive-NVFP4: Quantization
- philbert440/Qwen3.8-27B-Uncensored-Aggressive-MLX-4bit: Quantization
- inter521/Qwen3.8-27B-Uncensored-Aggressive-W4A16-AWQ: Quantization
- HFCK99/Qwen3.8-27B-Uncensored-Cyber: Abliterated