CCSSNE/tvall43-Qwen3.5-4B-heretic-v2
Abliterated vs Qwen/Qwen3.5-4B. Looks like abliteration (refusal removal): only the output projections (o_proj, out_proj) changed, in 29 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 0/196, Attention 29/320, Norms 0/203
2 derivatives checked
- mradermacher/tvall43-Qwen3.5-4B-heretic-v2-i1-GGUF: Quantization
- mradermacher/tvall43-Qwen3.5-4B-heretic-v2-GGUF: Same weights, converted
1 other repos hold byte-for-byte identical weight files.