MuXodious/LFM2.5-VL-1.6B-absolute-heresy-MPOA
Abliterated vs LiquidAI/LFM2.5-VL-1.6B. Looks like abliteration (refusal removal): only the output projections (out_proj, w2) changed, in 10 of 27 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/5, other 5/34, MLP 10/156, Norms 0/154, Attention 5/240
2 derivatives checked
- mradermacher/LFM2.5-VL-1.6B-absolute-heresy-MPOA-i1-GGUF: Quantization (fewer layers)
- mradermacher/LFM2.5-VL-1.6B-absolute-heresy-MPOA-GGUF: Same weights, converted (fewer layers)