prithivMLmods/TOI-157-Phi-4-Reasoning-Mini
Abliterated vs microsoft/Phi-4-mini-reasoning. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 32 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Norms 0/65, MLP 32/64, Attention 32/64
2 derivatives checked
- mradermacher/TOI-157-Phi-4-Reasoning-Mini-i1-GGUF: Quantization
- mradermacher/TOI-157-Phi-4-Reasoning-Mini-GGUF: Same weights, converted