dealignai/Qwen3.8-Flash-Next-UNCENSORED-FP8
Abliterated vs Qwen/Qwen3.8-Flash-Next-FP8. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 30 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/302, other 0/8, Attention 29/608, MLP 30/1943, Embeddings 0/138, lm_head 0/1
2 derivatives checked
- BattleGhost/Qwen3.8-Flash-Next-UNCENSORED-IQ2_XXS-GGUF: Quantization
- BattleGhost/Qwen3.8-Flash-Next-UNCENSORED-BF16-GGUF: Quantization
2 other repos hold byte-for-byte identical weight files.