MuXodious/Llama-3.3-8B-Instruct-128K-absolute-heresy
Abliterated vs shb777/Llama-3.3-8B-Instruct-128K. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 22 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/65, MLP 19/96, Attention 22/128, lm_head 0/1
2 derivatives checked
- mradermacher/Llama-3.3-8B-Instruct-128K-absolute-heresy-i1-GGUF: Quantization
- mradermacher/Llama-3.3-8B-Instruct-128K-absolute-heresy-GGUF: Same weights, converted