vulture3/G4-MeroWana-31B-heretic
Abliterated vs vulture3/G4-MeroWana-31B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 23 of 60 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/523, other 0/62, MLP 6/261, Attention 23/338
2 derivatives checked
- mradermacher/G4-MeroWana-31B-heretic-i1-GGUF: Quantization (fewer layers)
- mradermacher/G4-MeroWana-31B-heretic-GGUF: Quantization (fewer layers)