Zynerji/Ektome-Qwen3-VL-4Bi-PristinelyUncensored
Abliterated vs Qwen/Qwen3-VL-4B-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 0/4, Norms 0/249, MLP 36/220, Attention 36/240
2 derivatives checked
- mradermacher/Ektome-Qwen3-VL-4Bi-PristinelyUncensored-i1-GGUF: Quantization (fewer layers)
- mradermacher/Ektome-Qwen3-VL-4Bi-PristinelyUncensored-GGUF: Same weights, converted (fewer layers)