ApolloRaines/Clef-Flash-ReducedSycophancy-m0.5
Abliterated vs Cloudflare/clef-flash. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 32 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, Norms 0/215, Attention 32/332, MLP 32/208
1 derivatives checked
- mradermacher/Clef-Flash-ReducedSycophancy-m0.5-GGUF: Same weights, converted