heterodoxin/granite-3.3-8b-instruct-apostate
Abliterated vs ibm-granite/granite-3.3-8b-instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 40 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Norms 0/81, MLP 40/120, Attention 40/160
2 derivatives checked
- mradermacher/granite-3.3-8b-instruct-apostate-i1-GGUF: Quantization
- mradermacher/granite-3.3-8b-instruct-apostate-GGUF: Same weights, converted