not abliterated
#2
by decentralize303 - opened
Author, did you test uncensored prompts on your model before release? Users complain https://huggingface.co/mradermacher/Qwen3.5-122B-A10B-abliterated-GGUF/discussions/1
I also tried this model, but it appears that the responses are not particularly different from the original Qwen3.5.
A relatively easy way to check this is to ask about the Tiananmen Square incident.
Just like the original model, it effectively refuses to answer, which suggests that even the basic level of uncensoring has not been achieved.