Vikhrmodels/QVikhr-3-8B-Instruction
Abliterated vs Qwen/Qwen3-8B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 16 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 16/108, Attention 16/144, lm_head 0/1
3 derivatives checked
- Vikhrmodels/QVikhr-3-8B-Instruction-GGUF: Quantization
- mradermacher/QVikhr-3-8B-Instruction-i1-GGUF: Quantization
- mradermacher/QVikhr-3-8B-Instruction-GGUF: Same weights, converted