AEON-7/DFlash-Qwen3.5-27B-Uncensored
Abliterated vs Qwen/Qwen3.5-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 36 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 36/307, Attention 9/560, other 0/1, Norms 0/325
3 derivatives checked
- mradermacher/DFlash-Qwen3.5-27B-Uncensored-i1-GGUF: Quantization
- mradermacher/DFlash-Qwen3.5-27B-Uncensored-GGUF: Quantization
- AEON-7/DFlash-Qwen3.5-27B-Uncensored-NVFP4: Quantization