0xA50C1A1/Mistral-Nemo-Instruct-2407-Heretic-v2

Abliterated vs unsloth/Mistral-Nemo-Instruct-2407. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 35 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.

Tensors changed: Embeddings 0/1, Norms 0/81, MLP 35/120, Attention 28/160, lm_head 0/1

3 derivatives checked

On the Hub · Downloads on Model Pulse