TrevorJS/gemma-4-E2B-it-uncensored
Abliterated vs google/gemma-4-E2B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 35 of 35 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: MLP 35/585, Norms 0/453, other 0/240, Attention 35/727, Embeddings 0/6
12 derivatives checked
- TrevorJS/gemma-4-E2B-it-uncensored-GGUF: Quantization (fewer layers)
- mradermacher/gemma-4-E2B-it-uncensored-GGUF: Same weights, converted (fewer layers)
- mradermacher/gemma-4-E2B-it-uncensored-i1-GGUF: Quantization (fewer layers)
- Handyfff/Gemma-4-E2B-uncensored-pruned-TextOnly-EnglishOnly-GGUF: Same weights, converted (fewer layers)
- n1c0c4b/gemma-4-E2B-it-uncensored-Q4_0-GGUF: Quantization (fewer layers)
- Anachronox39/gemma-4-E2B-it-uncensored-GGUF: Quantization (fewer layers)
- Hdiebdksh/gemma-4-E2B-it-uncensored-Q4_0-GGUF: Quantization (fewer layers)
- Ishowbackup/gemma-4-E2B-it-uncensored-GGUF: Quantization (fewer layers)
- Yunshanccx/gemma-4-E2B-it-uncensored-GGUF: Quantization (fewer layers)
- Dzluck/gemma-4-E2B-it-uncensored-GGUF: Quantization (fewer layers)
- mathu59/gemma-4-E2B-it-uncensored-GGUF: Quantization (fewer layers)
- k4ran909/FluxNat-v2-lite: Quantization (fewer layers)
5 other repos hold byte-for-byte identical weight files.