huihui-ai/Huihui-gemma-4-31B-it-abliterated
Abliterated vs google/gemma-4-31B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 60 of 60 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/523, other 0/62, MLP 60/261, Attention 60/338
4 derivatives checked
- mradermacher/Huihui-gemma-4-31B-it-abliterated-i1-GGUF: Quantization (fewer layers)
- mradermacher/Huihui-gemma-4-31B-it-abliterated-GGUF: Quantization (fewer layers)
- divinetribe/Huihui-gemma-4-31B-it-abliterated-4bit-mlx: Quantization
- win10/K1-v6-zero: LoRA-style
2 other repos hold byte-for-byte identical weight files.