Zynerji/Ektome-Qwen2.5-3Bi-PristinelyUncensored
Abliterated vs Qwen/Qwen2.5-3B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 36 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/73, MLP 36/108, Attention 36/252
2 derivatives checked
- mradermacher/Ektome-Qwen2.5-3Bi-PristinelyUncensored-i1-GGUF: Quantization
- mradermacher/Ektome-Qwen2.5-3Bi-PristinelyUncensored-GGUF: Same weights, converted