heretic-org/Phi-4-mini-reasoning-heretic
Abliterated vs microsoft/Phi-4-mini-reasoning. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 25 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/65, MLP 19/64, Attention 25/64
1 derivatives checked
- SimplySara/Phi-4-mini-reasoning-heretic-i1-GGUF: Quantization