grimjim/Nemo-Instruct-2407-MPOA-v2-12B
Abliterated vs mistralai/Mistral-Nemo-Instruct-2407. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 25 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 25/120, Attention 25/160, lm_head 0/1
3 derivatives checked
- mradermacher/Nemo-Instruct-2407-MPOA-v2-12B-i1-GGUF: Quantization
- mradermacher/Nemo-Instruct-2407-MPOA-v2-12B-GGUF: Quantization
- grimjim/Nemo-Instruct-2407-baked-v1-12B: Abliterated