3MPER0RR/Ornith-1.5-35B-A3B-3MPER0RR-abliterated
Abliterated (fewer layers) vs huihui-ai/Huihui-Ornith-1.5-35B-A3B-abliterated. 3MPER0RR/Ornith-1.5-35B-A3B-3MPER0RR-abliterated matches part of huihui-ai/Huihui-Ornith-1.5-35B-A3B-abliterated (1026 of its 1811 tensors; the rest of huihui-ai/Huihui-Ornith-1.5-35B-A3B-abliterated has no counterpart). On the shared part: Looks like abliteration (refusal removal): only the output projections (o_proj, out_proj) changed, in 5 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/241, Attention 5/388, MLP 0/392, lm_head 0/1