Auguments/Aureth-4B-Qwen3.5-Heretic
Abliterated vs OusiaResearch/Aureth-4B-Qwen3.5. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 28 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/203, Attention 28/320, MLP 7/196
2 derivatives checked
- mradermacher/Aureth-4B-Qwen3.5-Heretic-i1-GGUF: Quantization
- mradermacher/Aureth-4B-Qwen3.5-Heretic-GGUF: Same weights, converted