saidutta69/MiMo-V2.6-Distill-Qwen-9B-heretic
Abliterated vs XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B. 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: lm_head 0/1, Embeddings 0/4, MLP 19/208, Attention 28/332, Norms 0/215
2 derivatives checked
- mradermacher/MiMo-V2.6-Distill-Qwen-9B-heretic-i1-GGUF: Quantization
- mradermacher/MiMo-V2.6-Distill-Qwen-9B-heretic-GGUF: Same weights, converted