temaq-org/Tema_Q-X2-4B-Thinking
Abliterated (fewer layers) vs Qwen/Qwen3.5-4B. temaq-org/Tema_Q-X2-4B-Thinking matches part of Qwen/Qwen3.5-4B (426 of its 738 tensors; the rest of Qwen/Qwen3.5-4B has no counterpart). On the shared part: Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 19 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, MLP 19/96, Attention 11/224, Norms 0/105
2 derivatives checked
- mradermacher/Tema_Q-X2-4B-Thinking-i1-GGUF: Quantization
- mradermacher/Tema_Q-X2-4B-Thinking-GGUF: Same weights, converted