DavidAU/Qwen3-4B-Instruct-2507-Polaris-Alpha-Distill-Heretic-Abliterated
Abliterated vs TeichAI/Qwen3-4B-Instruct-2507-Polaris-Alpha-Distill. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 33 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 8/108, Attention 33/144
3 derivatives checked
- mradermacher/Qwen3-4B-Instruct-2507-Polaris-Alpha-Distill-Heretic-Abliterated-GGUF: Same weights, converted
- mradermacher/Qwen3-4B-Instruct-2507-Polaris-Alpha-Distill-Heretic-Abliterated-i1-GGUF: Quantization
- ZeroXClem/Qwen3-4B-Sky-High-Hermes: Full fine-tune