Kilinskiy/Step-3.5-Flash-Ablitirated

Abliterated vs stepfun-ai/Step-3.5-Flash. Looks like abliteration (refusal removal): only the output projections (down_proj) changed, in 4 of 45 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.

Tensors changed: other 0/84, Embeddings 0/1, Norms 0/181, MLP 4/261, Attention 0/225, lm_head 0/1

2 derivatives checked

On the Hub · Downloads on Model Pulse