heswithme/Huihui-Qwen3.8-27B-Abliterated-Gittensor-Style-NVFP4-RTX5090
Abliterated vs gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 62 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Attention 132/1180, MLP 152/880, Embeddings 0/4, Norms 0/319, lm_head 0/4