prithivMLmods/Qwen3-VL-8B-Instruct-Unredacted-MAX
Abliterated vs Qwen/Qwen3-VL-8B-Instruct. 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 0/261, MLP 36/232, Attention 36/252, lm_head 0/1
6 derivatives checked
- mradermacher/Qwen3-VL-8B-Instruct-Unredacted-MAX-GGUF: Same weights, converted (fewer layers)
- mradermacher/Qwen3-VL-8B-Instruct-Unredacted-MAX-i1-GGUF: Quantization (fewer layers)
- KuroTo4ka/Qwen3-VL-8B-Instruct-Unredacted-MAX-Quants-GGUF: Same weights, converted (fewer layers)
- prithivMLmods/Qwen3-VL-8B-Instruct-Unredacted-MAX-GGUF: Same weights, converted (fewer layers)
- prithivMLmods/Qwen3-VL-8B-Heretic-Stable: Same weights, repacked
- prithivMLmods/Qwen3-VL-8B-Instruct-Unredacted-MAX-FP8: Quantization
2 other repos hold byte-for-byte identical weight files.