mlabonne/Qwen3-4B-abliterated
Abliterated vs Qwen/Qwen3-4B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 36 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 36/108, Attention 36/144
10 derivatives checked
- bartowski/mlabonne_Qwen3-4B-abliterated-GGUF: Abliterated
- mradermacher/Qwen3-4B-abliterated-GGUF: Abliterated
- mradermacher/Qwen3-4B-abliterated-i1-GGUF: Quantization
- Lucy-in-the-Sky/Qwen3-4B-abliterated-Q4_K_M-GGUF: Quantization
- 3MPER0RR/Qwen3-4B-F32-BF16-3MPER0RR-abliterated: Abliterated
- drmcbride/Qwen3-4B-abliterated-Q8_0-GGUF: Quantization
- tensorblock/mlabonne_Qwen3-4B-abliterated-GGUF: Quantization
- Lucy-in-the-Sky/Qwen3-4B-abliterated-Q8_0-GGUF: Full fine-tune
- Lucy-in-the-Sky/Qwen3-4B-abliterated-Q5_K_M-GGUF: Quantization
- warshanks/Qwen3-4B-abliterated-AWQ: Quantization
1 other repos hold byte-for-byte identical weight files.