coder3101/Qwen3-VL-8B-Thinking-heretic
Abliterated vs Qwen/Qwen3-VL-8B-Thinking. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 22 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/261, MLP 22/232, Attention 9/252, lm_head 0/1
2 derivatives checked
- chenly124/Qwen3-VL-8B-Thinking-heretic-Q4_K_M-GGUF: Quantization (fewer layers)
- Music4Dogs/Qwen3-VL-8B-Thinking-heretic-Q5_K_M-GGUF: Quantization (fewer layers)