tvall43/Qwen3.5-2B-Claude-4.6-Opus-Reasoning-Distilled-heretic-v1
Abliterated vs Jackrong/Qwen3.5-2B-Claude-4.6-Opus-Reasoning-Distilled. Looks like abliteration (refusal removal): only the output projections (o_proj, out_proj) changed, in 19 of 24 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Attention 19/264, Norms 0/177, Embeddings 0/4, MLP 0/172
2 derivatives checked
- mradermacher/Qwen3.5-2B-Claude-4.6-Opus-Reasoning-Distilled-heretic-v1-i1-GGUF: Quantization
- mradermacher/Qwen3.5-2B-Claude-4.6-Opus-Reasoning-Distilled-heretic-v1-GGUF: Same weights, converted