DreamFast/Qwen3-VL-4b-Heretic-GGUF
Abliterated (fewer layers) vs Qwen/Qwen3-VL-4B-Instruct. DreamFast/Qwen3-VL-4b-Heretic-GGUF matches part of Qwen/Qwen3-VL-4B-Instruct (398 of its 713 tensors; the rest of Qwen/Qwen3-VL-4B-Instruct has no counterpart). On the shared part: DreamFast/Qwen3-VL-4b-Heretic-GGUF is a GGUF (F16) of different weights. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 32 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 30/108, Attention 32/144