prithivMLmods/VibeThinker-3B-heretic_decensored
Abliterated vs WeiboAI/VibeThinker-3B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 34 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/73, MLP 25/108, Attention 34/252
3 derivatives checked
- mradermacher/VibeThinker-3B-heretic_decensored-i1-GGUF: Quantization
- prithivMLmods/VibeThinker-3B-heretic_decensored-GGUF: Same weights, converted
- mradermacher/VibeThinker-3B-heretic_decensored-GGUF: Same weights, converted