UnstableLlama/Laguna-S-2.1-Heretic
Abliterated vs poolside/Laguna-S-2.1. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 23 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/193, MLP 23/2565, Attention 23/240
2 derivatives checked
- mradermacher/Laguna-S-2.1-Heretic-i1-GGUF: Quantization
- mradermacher/Laguna-S-2.1-Heretic-GGUF: Quantization