MuXodious/Llama-3.3-8B-Instruct-128K-PaperWitch-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 32 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 32/96, Attention 31/128, lm_head 0/1
3 derivatives checked
- mradermacher/Llama-3.3-8B-Instruct-128K-PaperWitch-heresy-i1-GGUF: Quantization
- mradermacher/Llama-3.3-8B-Instruct-128K-PaperWitch-heresy-GGUF: Same weights, converted
- N-Bot-Int/MrgrtV1-8B-merged: LoRA-style