ChiKoi7/Falcon3-7B-Instruct-Heretic-v2
Abliterated vs tiiuae/Falcon3-7B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 21 of 28 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/57, MLP 19/84, Attention 21/112, lm_head 0/1
2 derivatives checked
- mradermacher/Falcon3-7B-Instruct-Heretic-v2-i1-GGUF: Quantization
- mradermacher/Falcon3-7B-Instruct-Heretic-v2-GGUF: Same weights, converted