huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2
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
10 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-GGUF: Quantization
- DevQuasar-7/huihui-ai.DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-GGUF: Same weights, converted
- mradermacher/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-i1-GGUF: Quantization
- TheWirelessPhoenix/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-mlx_4bit: Quantization
- tensorblock/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-GGUF: Quantization
- asa123ss/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-Q4_K_M-GGUF: Quantization
- lhca521/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-GGUF: Quantization
- mlx-community/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-Q4-mlx: Quantization
- FiditeNemini/Unhinged-Qwen2.5-R1-1M-Uncensored-BF16: Abliterated
- FiditeNemini/Qwen2.5-14B-DeepSeek-R1-1M-Uncensored: Abliterated
3 other repos hold byte-for-byte identical weight files.