prithivMLmods/Gliese-Qwen3.5-0.8B-Abliterated-Caption
Abliterated vs Qwen/Qwen3.5-0.8B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 24 of 24 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Attention 24/216, Norms 0/129, Embeddings 0/4, MLP 24/124
2 derivatives checked
- mradermacher/Gliese-Qwen3.5-0.8B-Abliterated-Caption-i1-GGUF: Quantization
- mradermacher/Gliese-Qwen3.5-0.8B-Abliterated-Caption-GGUF: Same weights, converted