ZuzeTt/Qwen3-VL-4B-Thinking-heretic-imatrix-GGUF

Abliterated (fewer layers) vs Qwen/Qwen3-VL-4B-Thinking. ZuzeTt/Qwen3-VL-4B-Thinking-heretic-imatrix-GGUF matches part of Qwen/Qwen3-VL-4B-Thinking (398 of its 713 tensors; the rest of Qwen/Qwen3-VL-4B-Thinking has no counterpart). On the shared part: ZuzeTt/Qwen3-VL-4B-Thinking-heretic-imatrix-GGUF is a GGUF (BF16) 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 32/108, Attention 31/144

On the Hub · Downloads on Model Pulse