spitfire4794/LFM2.5-1.2B-Instruct-Heretic
Abliterated vs LiquidAI/LFM2.5-1.2B-Instruct. Looks like abliteration (refusal removal): only the output projections (out_proj, w2) changed, in 14 of 16 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/2, other 9/30, MLP 13/48, Norms 0/44, Attention 0/24
2 derivatives checked
- mradermacher/LFM2.5-1.2B-Instruct-Heretic-i1-GGUF: Quantization
- mradermacher/LFM2.5-1.2B-Instruct-Heretic-GGUF: Same weights, converted