prithivMLmods/Qwen3.5-0.8B-Unredacted-MAX
Abliterated vs Qwen/Qwen3.5-0.8B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 24 of 24 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Attention 24/216, Norms 0/129, Embeddings 0/4, MLP 24/124
4 derivatives checked
- mradermacher/Qwen3.5-0.8B-Unredacted-MAX-i1-GGUF: Quantization
- prithivMLmods/Qwen3.5-Thinking-AIO-GGUF: Same weights, converted
- mradermacher/Qwen3.5-0.8B-Unredacted-MAX-GGUF: Same weights, converted
- machinadeusex/Qwen3.5-Thinking-AIO-GGUF: Same weights, converted