koute/GLM-4.7-Flash-Derestricted
Abliterated vs zai-org/GLM-4.7-Flash. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 35 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/2, Norms 0/196, MLP 599/2559, Attention 35/240, lm_head 0/1, other 0/2
2 derivatives checked
- mradermacher/GLM-4.7-Flash-Derestricted-GGUF: Quantization
- mradermacher/GLM-4.7-Flash-Derestricted-i1-GGUF: Quantization