collinzrj/DeepSeek-R1-Distill-Llama-8B-abliterate
Abliterated vs deepseek-ai/DeepSeek-R1-Distill-Llama-8B. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 32 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Attention 32/128, MLP 32/96, Norms 0/65, lm_head 0/1
2 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Llama-8B-abliterate-i1-GGUF: Quantization
- mradermacher/DeepSeek-R1-Distill-Llama-8B-abliterate-GGUF: Same weights, converted