spitfire4794/LFM2-700M-Heretic
Abliterated vs LiquidAI/LFM2-700M. Looks like abliteration (refusal removal): only the output projections (out_proj, w2) changed, in 10 of 16 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/2, other 4/30, MLP 10/48, Norms 0/44, Attention 0/24
2 derivatives checked
- mradermacher/LFM2-700M-Heretic-i1-GGUF: Quantization
- mradermacher/LFM2-700M-Heretic-GGUF: Same weights, converted