llmfan46/gemma-4-31B-it-qat-q4_0-unquantized-uncensored-heretic
Abliterated vs google/gemma-4-31B-it-qat-q4_0-unquantized. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 26 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 0/261, Attention 26/338
9 derivatives checked
- mradermacher/gemma-4-31B-it-qat-q4_0-unquantized-uncensored-heretic-i1-GGUF: Quantization (fewer layers)
- llmfan46/gemma-4-31B-it-qat-q4_0-uncensored-heretic-GGUF: Quantization (fewer layers)
- EZForever/gemma-4-31B-it-qat-uncensored-heretic-UDmerge-GGUF: Quantization (fewer layers)
- ailexleon/gemma-4-31B-it-qat-uncensored-heretic-mlx-lm-4Bit: Quantization
- mradermacher/gemma-4-31B-it-qat-q4_0-unquantized-uncensored-heretic-GGUF: Quantization (fewer layers)
- llmfan46/gemma-4-31B-it-qat-q4_0-uncensored-heretic-NVFP4-GGUF: Same weights, converted
- ailexleon/gemma-4-31B-it-qat-uncensored-heretic-mlx-vlm-4Bit: Quantization
- llmfan46/gemma-4-31B-it-qat-q4_0-uncensored-heretic-NVFP4: Quantization
- llmfan46/gemma-4-31B-it-qat-q4_0-uncensored-heretic-GPTQ-Int4: Quantization