SOMEHOTMEAL/Qwen3.5-9B-Janus-Abliterated-V2-Safetensors
Abliterated (fewer layers) vs Qwen/Qwen3.5-9B. SOMEHOTMEAL/Qwen3.5-9B-Janus-Abliterated-V2-Safetensors matches part of Qwen/Qwen3.5-9B (439 of its 775 tensors; the rest of Qwen/Qwen3.5-9B has no counterpart). On the shared part: Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 31 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: MLP 31/99, Attention 31/228, Norms 0/111, Embeddings 0/1
2 derivatives checked
- mradermacher/Qwen3.5-9B-Janus-Abliterated-V2-Safetensors-i1-GGUF: Quantization
- mradermacher/Qwen3.5-9B-Janus-Abliterated-V2-Safetensors-GGUF: Same weights, converted