ChiKoi7/Olmo-3-7B-Instruct-Heretic
Abliterated vs allenai/Olmo-3-7B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 27 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, MLP 27/96, Norms 0/129, Attention 26/128, lm_head 0/1
2 derivatives checked
- mradermacher/Olmo-3-7B-Instruct-Heretic-i1-GGUF: Quantization
- mradermacher/Olmo-3-7B-Instruct-Heretic-GGUF: Same weights, converted