thirdeyeai/DeepSeek-R1-Distill-Qwen-1.5B-uncensored
Abliterated vs deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, o_proj) changed, in 28 of 28 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 1/1, Attention 28/196, MLP 28/84, Norms 0/57, lm_head 0/1
2 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Qwen-1.5B-uncensored-GGUF: Same weights, converted
- mradermacher/DeepSeek-R1-Distill-Qwen-1.5B-uncensored-i1-GGUF: Quantization
2 other repos hold byte-for-byte identical weight files.