huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated
Abliterated vs deepseek-ai/DeepSeek-R1-Distill-Qwen-32B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 63 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Attention 63/448, MLP 63/192, Norms 0/129, lm_head 0/1
12 derivatives checked
- bartowski/DeepSeek-R1-Distill-Qwen-32B-abliterated-GGUF: Same weights, converted
- Thaurock/DeepSeek-R1-Distill-Qwen-32B-abliterated-GGUF: Same weights, converted
- mradermacher/DeepSeek-R1-Distill-Qwen-32B-abliterated-i1-GGUF: Quantization
- mradermacher/DeepSeek-R1-Distill-Qwen-32B-abliterated-GGUF: Quantization
- mlx-community/DeepSeek-R1-Distill-Qwen-32B-abliterated-4bit: Quantization
- DevQuasar-4/huihui-ai.DeepSeek-R1-Distill-Qwen-32B-abliterated-GGUF: Quantization
- lactroiii/DeepSeek-R1-Distill-Qwen-32B-abliterated-GGUF: Quantization
- S010/DeepSeek-R1-Distill-Qwen-32B-abliterated-Q8-mlx: Quantization
- mlx-community/DeepSeek-R1-Distill-Qwen-32B-abliterated: Same weights, converted
- DavidAU/Qwen2.5-QwQ-35B-Eureka-Cubed-abliterated-uncensored: Full fine-tune
- DavidAU/Qwen2.5-QwQ-37B-Eureka-Triple-Cubed-abliterated-uncensored: Full fine-tune
- DataSoul/DwQ-R1-32B-v0.1: Full fine-tune
2 other repos hold byte-for-byte identical weight files.