Blackfrost-AI/Muse-Glimmer-30B-Abliterated-BF16
Abliterated vs meta-models/Muse-Glimmer-30B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 50 of 52 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/3, Norms 0/413, MLP 50/358, Attention 50/660, lm_head 0/1, other 0/1
10 derivatives checked
- The-freezer/Muse-Glimmer-30B-Abliterated-GGUF: Full fine-tune
- mradermacher/Muse-Glimmer-30B-Abliterated-BF16-i1-GGUF: Quantization
- mradermacher/Muse-Glimmer-30B-Abliterated-BF16-GGUF: Quantization
- 0xSojalSec/Muse-Glimmer-30B-Abliterated-GGUF: Full fine-tune
- Blackfrost-AI/Muse-Glimmer-30B-Abliterated-NVFP4: Quantization
- shoemoney/Muse-Glimmer-30B-Abliterated-MLX-q6: Quantization
- Ishowbackup/Muse-Glimmer-30B-Abliterated-GGUF: Full fine-tune
- shoemoney/Muse-Glimmer-30B-Abliterated-MLX-q8: Quantization
- shoemoney/Muse-Glimmer-30B-Abliterated-MLX-mixed36: Quantization
- Ishowbackup/Muse-Glimmer-30B-Abliterated-MLX-4bit: Quantization
1 other repos hold byte-for-byte identical weight files.