byroneverson/Mistral-Small-Instruct-2409-abliterated
Abliterated vs mistralai/Mistral-Small-Instruct-2409. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 44 of 56 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Norms 0/113, MLP 31/168, Attention 28/224, lm_head 0/1