News for hackers. Stuff that matters.

A comment on DeepSeek releases Janus Pro, a text-to-image generator [pdf]. Parent comment

Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by porphyra on Monday January 27, 2025 @02:05PM (#42844479)
Why did you misspell Tiananmen (composed of three characters "Tian An Men") three times? There's an 'n' after the 'a'.
[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by razster on Monday January 27, 2025 @02:22PM (#42844628)
Even my local distilled models, 32b and 14 refuse to answer, even my escape prompt is met with the same reply that it cannot provide that answer.
[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by TuxSH on Monday January 27, 2025 @02:56PM (#42844968)
For the 14b model, you can "/set parameter temperature 0.4" to make the Chinese censorship go away

Ah and you need to ask it to answer factually, too. Actually, asking it to answer factually does remove a lot of the censorship by itself.

[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by razster on Monday January 27, 2025 @02:25PM (#42844661)
Seems there are a bunch of Uncensored models out there, going to give them a run and see.
[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by KennyBlanken on Monday January 27, 2025 @02:23PM (#42844636)
Sometimes misspelling things causes the naughty-no-no filters to miss the input, but the LLM indentifies it correctly.
[ Parent | on HN ]