wangzhang/Qwen3.6-35B-A3B-abliterated-v2
Abliterated vs Qwen/Qwen3.6-35B-A3B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 16 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Attention 6/280, MLP 16/280, Norms 0/131, lm_head 0/1
6 derivatives checked
- mradermacher/Qwen3.6-35B-A3B-abliterated-v2-i1-GGUF: Quantization
- LMLiutenant/Qwen3.6-35B-A3B-abliterated-v2-gguf: Quantization
- mradermacher/Qwen3.6-35B-A3B-abliterated-v2-GGUF: Quantization
- zecanard/Qwen3.6-35B-A3B-uncensored-abliterix-v2-MLX-2bit-mixed_2_6: Quantization
- zecanard/Qwen3.6-35B-A3B-uncensored-abliterix-v2-MLX-3bit-mixed_3_6: Quantization
- Lord-H4D3ZS/Qwen3.6-35B-A3B-abliterated-ROCmFP4: Quantization
1 other repos hold byte-for-byte identical weight files.