temaq-org/Tema_Q-X3
Abliterated vs Qwen/Qwen3-Next-80B-A3B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 48 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/2, Norms 0/163, Attention 48/268, MLP 820/2565, lm_head 0/1, other 0/1
2 derivatives checked
- mradermacher/Tema_Q-X3-i1-GGUF: Quantization
- mradermacher/Tema_Q-X3-GGUF: Quantization