huihui-ai/Huihui-LFM2.5-1.2B-Instruct-abliterated
Abliterated vs LiquidAI/LFM2.5-1.2B-Instruct. Looks like abliteration (refusal removal): only the output projections (out_proj, w2) changed, in 16 of 16 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/2, other 10/30, MLP 16/48, Norms 0/44, Attention 6/24
4 derivatives checked
- mradermacher/Huihui-LFM2.5-1.2B-Instruct-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-LFM2.5-1.2B-Instruct-abliterated-GGUF: Same weights, converted
- Iambackup/Huihui-LFM2.5-1.2B-Instruct-abliterated-GGUF: Same weights, converted
- blascotobasco/LFM2.5-4E-A2.4B-Abliterated: LoRA-style
1 other repos hold byte-for-byte identical weight files.