coder3101/gemma-4-E2B-it-heretic
Abliterated vs google/gemma-4-E2B-it. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 9 of 35 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: MLP 0/585, Norms 0/453, other 0/240, Attention 9/727, Embeddings 0/6
5 derivatives checked
- mradermacher/gemma-4-E2B-it-heretic-i1-GGUF: Quantization (fewer layers)
- JMingo/gemma-4-E2B-it-heretic-UD-GGUF: Same weights, converted (fewer layers)
- mradermacher/gemma-4-E2B-it-heretic-GGUF: Same weights, converted (fewer layers)
- L-Alchemyst/Coder3101-Gemma-4-E2B-it-heretic-GGUF: Same weights, converted (fewer layers)
- Stabhappy/gemma-4-E2B-it-heretic-GGUF: Quantization (fewer layers)