MuXodious/LFM2-8B-A1B-absolute-heresy-MPOA
Abliterated vs LiquidAI/LFM2-8B-A1B. Looks like abliteration (refusal removal): only the output projections (out_proj) changed, in 17 of 24 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: MLP 0/2162, Embeddings 0/2, other 13/54, Norms 0/60, Attention 4/24
3 derivatives checked
- mradermacher/LFM2-8B-A1B-absolute-heresy-MPOA-i1-GGUF: Quantization
- mradermacher/LFM2-8B-A1B-absolute-heresy-MPOA-GGUF: Same weights, converted
- blascotobasco/LFM2-16B-A2B: LoRA-style (fewer layers)