DreamFast/Qwen3-VL-4b-Heretic
Abliterated vs Qwen/Qwen3-VL-4B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 32 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 30/220, Attention 32/240
2 derivatives checked
- mradermacher/Qwen3-VL-4b-Heretic-GGUF: Same weights, converted (fewer layers)
- mradermacher/Qwen3-VL-4b-Heretic-i1-GGUF: Quantization (fewer layers)