Vortex5/Pantheon-Reasoning-26B-A4B-1.1-heretic
Abliterated vs Gryphe/Pantheon-Reasoning-26B-A4B-1.1. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 23 of 30 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 0/231, Norms 0/433, other 0/122, Attention 23/223, lm_head 0/1
3 derivatives checked
- mradermacher/Pantheon-Reasoning-26B-A4B-1.1-heretic-i1-GGUF: Quantization
- mradermacher/Pantheon-Reasoning-26B-A4B-1.1-heretic-GGUF: Quantization
- SubMaroon/Kitchoon-26B-A4B: LoRA-style