3MPER0RR/DeepSeek-R1-Distill-Llama-8B-3MPER0RR-abliterated
Abliterated vs deepseek-ai/DeepSeek-R1-Distill-Llama-8B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 26 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Attention 25/128, MLP 26/96, Norms 0/65, lm_head 0/1
3 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Llama-8B-3MPER0RR-abliterated-i1-GGUF: Quantization
- mradermacher/DeepSeek-R1-Distill-Llama-8B-3MPER0RR-abliterated-GGUF: Same weights, converted
- 3MPER0RR/DeepSeek-R1-Distill-Llama-8B-3MPER0RR-abliterated-Q8_0-GGUF: Quantization