cesun/ThinkEdit-deepseek-llama3-8b
Abliterated vs deepseek-ai/DeepSeek-R1-Distill-Llama-8B. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 8 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Attention 8/128, MLP 0/96, Norms 0/65, lm_head 0/1
2 derivatives checked
- mradermacher/ThinkEdit-deepseek-llama3-8b-i1-GGUF: Quantization
- mradermacher/ThinkEdit-deepseek-llama3-8b-GGUF: Abliterated