Search Results
7/9/2025, 10:59:25 PM
>>102094853
I use R1 now but I noticed in the past that V3 can be wrangled similarly.
I had a similar issue where it would always end its messages with a question or statement like "What will you do?" or "Your move." I found the best thing to do is just to not overthink it, these models are pretty good at following instructions, though they get worse with bigger contexts.
I just added this sentence to my prompt and it just went away (though some of its endings are still shitty, it likes to explain that characters are "Waiting. Trembling." or whatever).
IMO, DeepSeek's strength lies in its uncensoredness. It means you're not hunting for magic phrases or trying to evade filters, you can just tell it what you want like an idiot.
I use R1 now but I noticed in the past that V3 can be wrangled similarly.
I had a similar issue where it would always end its messages with a question or statement like "What will you do?" or "Your move." I found the best thing to do is just to not overthink it, these models are pretty good at following instructions, though they get worse with bigger contexts.
I just added this sentence to my prompt and it just went away (though some of its endings are still shitty, it likes to explain that characters are "Waiting. Trembling." or whatever).
IMO, DeepSeek's strength lies in its uncensoredness. It means you're not hunting for magic phrases or trying to evade filters, you can just tell it what you want like an idiot.
Page 1