prithivMLmods/Polaris-VGA-27B-Post1.0e
Abliterated vs Qwen/Qwen3.5-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 64 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, MLP 64/304, Attention 64/556, Norms 0/319
2 derivatives checked
- mradermacher/Polaris-VGA-27B-Post1.0e-i1-GGUF: Quantization
- mradermacher/Polaris-VGA-27B-Post1.0e-GGUF: Quantization