huihui-ai/Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated
Abliterated vs Qwen/Qwen3-Coder-30B-A3B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 48 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/193, MLP 838/2613, Attention 48/192, lm_head 0/1
9 derivatives checked
- mradermacher/Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-GGUF: Quantization
- SAPSAN-SKLEP/HIDra-30B-A3B-GGUF-uncensored-cybersec: Full fine-tune
- Thaurock/Qwen3-Coder-30B-A3B-Instruct-abliterated-GGUF: Same weights, converted
- Johnblick187/Qwen3-Omni-Coder: Full fine-tune (fewer layers)
- marafx2007/Huihui-Qwen3-Coder-30B-A3B-Instruct-abliterated-i1-GGUF: Quantization
- bostonstrong567/Luau-Qwen3-Coder-30B-A3B: LoRA-style
- rolex0775/Qwen3-Coder-30B-A3B-Instruct-DQ-abliterated-GGUF: Quantization
- Tsotchke-Corporation/Qwen3-Coder-30B-A3B-GeoRefine-C30-abl-GGUF: Quantization
3 other repos hold byte-for-byte identical weight files.