huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated
Abliterated vs deepseek-ai/DeepSeek-R1-Distill-Qwen-14B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 47 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Attention 47/336, MLP 47/144, Norms 0/97, lm_head 0/1
4 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Qwen-14B-abliterated-i1-GGUF: Quantization
- mradermacher/DeepSeek-R1-Distill-Qwen-14B-abliterated-GGUF: Quantization
- CuckmeisterFuller/DeepSeek-R1-Distill-Qwen-14B-abliterated-Q4-mlx: Quantization
- tensorblock/DeepSeek-R1-Distill-Qwen-14B-abliterated-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.