DavidAU/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-Distill-Heretic-Abliterated
Abliterated vs TeichAI/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-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 15/108, Attention 33/144
2 derivatives checked
- mradermacher/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-Distill-Heretic-Abliterated-i1-GGUF: Quantization
- mradermacher/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-Distill-Heretic-Abliterated-GGUF: Same weights, converted
1 other repos hold byte-for-byte identical weight files.