huihui-ai/Huihui-LFM2-24B-A2B-abliterated
Abliterated vs LiquidAI/LFM2-24B-A2B. Looks like abliteration (refusal removal): only the output projections (out_proj, w2) changed, in 40 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: MLP 904/2768, Embeddings 0/2, other 30/90, Norms 0/100, Attention 10/40
3 derivatives checked
- mradermacher/Huihui-LFM2-24B-A2B-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-LFM2-24B-A2B-abliterated-GGUF: Quantization
- blascotobasco/LFM2-24B-A2B-CinderWatch: Partial fine-tune