huihui-ai/Llama-3.3-70B-Instruct-abliterated
Abliterated vs meta-llama/Llama-3.3-70B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 79 of 80 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/161, MLP 79/240, Attention 79/320, lm_head 0/1
19 derivatives checked
- bartowski/Llama-3.3-70B-Instruct-abliterated-GGUF: Quantization
- mradermacher/Llama-3.3-70B-Instruct-abliterated-GGUF: Quantization
- divinetribe/Llama-3.3-70B-Instruct-abliterated-8bit-mlx: Quantization
- jiangchengchengNLP/llama3.3-70B-instruct-abliterated-awq: Quantization
- mradermacher/Llama-3.3-70B-Instruct-abliterated-i1-GGUF: Quantization
- Azerty76/Llama-3.3-70B-Instruct-abliterated-i1-GGUF: Quantization
- Justbackup/Llama-3.3-70B-Instruct-abliterated-i1-GGUF: Quantization
- Justbackup/Llama-3.3-70B-Instruct-abliterated-GGUF: Quantization
- frscrcc/Llama-3.3-70B-Instruct-abliterated-mlx-4Bit: Quantization
- cnfusion/Llama-3.3-70B-Instruct-abliterated-Q2-mlx: Quantization
- tensorblock/Llama-3.3-70B-Instruct-abliterated-GGUF: Quantization
- schonsense/70B_Book_stock: LoRA-style
- Nexesenex/Llama_3.x_70b_DoppelGutenberg-L3.3_abliterated_fusion: Full fine-tune
- Nexesenex/Llama_3.x_70b_Nemotron-L3.3_abliterated_fusion_v2: Full fine-tune
- Nexesenex/Llama_3.x_70b_L3.3-Nemotron_abliterated_fusion: Partial fine-tune
- Nexesenex/Llama_3.x_70b_DoppelGutenberg-L3.3_abliterated_fusion_v2: Full fine-tune
- Nexesenex/Llama_3.x_70b_L3.3-Tulu-3_abliterated_fusion: Partial fine-tune
- Nexesenex/Llama_3.x_70b_L3.3-DoppelGutenberg_abliterated_fusion: Partial fine-tune
- Shifusen/L3.3-70B-PippaMaid-1.0: LoRA-style
6 other repos hold byte-for-byte identical weight files.