llmfan46/Q3.5-BlueStar-v2-27B-uncensored-heretic
Abliterated vs zerofata/Q3.5-BlueStar-v2-27B. Looks like abliteration (refusal removal): only the output projections (o_proj, out_proj) changed, in 42 of 64 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 0/319, Attention 42/556, MLP 0/304
3 derivatives checked
- mradermacher/Q3.5-BlueStar-v2-27B-uncensored-heretic-i1-GGUF: Quantization
- mradermacher/Q3.5-BlueStar-v2-27B-uncensored-heretic-GGUF: Quantization
- llmfan46/Q3.5-BlueStar-v2-27B-uncensored-heretic-GGUF: Same weights, converted