prithivMLmods/Qwen3-VL-2B-Instruct-abliterated-v1
Abliterated vs Qwen/Qwen3-VL-2B-Instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 28 of 28 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/217, MLP 28/196, Attention 28/208
6 derivatives checked
- prithivMLmods/Qwen3-VL-2B-Instruct-abliterated-v1-GGUF: Same weights, converted (fewer layers)
- mradermacher/Qwen3-VL-2B-Instruct-abliterated-i1-GGUF: Quantization (fewer layers)
- mradermacher/Qwen3-VL-2B-Instruct-abliterated-GGUF: Abliterated (fewer layers)
- mradermacher/Qwen3-VL-2B-Instruct-abliterated-v1-i1-GGUF: Quantization (fewer layers)
- mradermacher/Qwen3-VL-2B-Instruct-abliterated-v1-GGUF: Same weights, converted (fewer layers)
- prithivMLmods/Qwen3-VisionCaption-2B: LoRA-style