prithivMLmods/Qwen3-VL-4B-Thinking-Unredacted-MAX
Abliterated vs Qwen/Qwen3-VL-4B-Thinking. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 36 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/249, MLP 36/220, Attention 36/240
3 derivatives checked
- mradermacher/Qwen3-VL-4B-Thinking-Unredacted-MAX-i1-GGUF: Quantization (fewer layers)
- mradermacher/Qwen3-VL-4B-Thinking-Unredacted-MAX-GGUF: Same weights, converted (fewer layers)
- prithivMLmods/Qwen3-VL-4B-Thinking-Unredacted-MAX-GGUF: Same weights, converted (fewer layers)
1 other repos hold byte-for-byte identical weight files.