huihui-ai/Huihui-Qwen3-VL-8B-Instruct-abliterated
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
9 derivatives checked
- mradermacher/Huihui-Qwen3-VL-8B-Instruct-abliterated-GGUF: Same weights, converted (fewer layers)
- noctrex/Huihui-Qwen3-VL-8B-Instruct-abliterated-GGUF: Same weights, converted (fewer layers)
- mradermacher/Huihui-Qwen3-VL-8B-Instruct-abliterated-i1-GGUF: Quantization (fewer layers)
- noctrex/Huihui-Qwen3-VL-8B-Instruct-abliterated-i1-GGUF: Quantization (fewer layers)
- porpgel/Huihui-Qwen3-VL-8B-Instruct-abliterated-Q4_K_M-GGUF: Quantization (fewer layers)
- quant-mind/Huihui-Qwen3-VL-8B-Instruct-abliterated-AWQ-int4: Quantization
- lihaoyun6/Qwen3-VL-8B-Instruct-abliterated_GGUF: Quantization (fewer layers)
- Schwi07/Huihui-Qwen3-VL-8B-Instruct-abliterated-Q6_K-GGUF: Quantization (fewer layers)
- Heouzen/Huihui-Qwen3-VL-8B-Instruct-FP8-abliterated: Partial fine-tune
7 other repos hold byte-for-byte identical weight files.