Not adequate settings in compare to answers
I had set in the chat settings that the answers should be as much precise possible, with margin that wouldn’t limit chat’s creativity.
Unfortunatelly, while answering, the chat hasn’t strictly followed the instructions, had copied (or re-wrote) same answer, despite of my corrections (with caused me me loose tokens for them)
Log in to comment and vote
Comments16
Jack
Jun 11, 2025
Which model are you seeing this with? And what is your temperature and top_p set to?
Amethyst Lynx
Jun 12, 2025
Hello. I use the defalt set model and the other paraneters are set that they would give much precise answer, while remaining creative.
The parameter in with the max is 1, is set to around 0.7.
Amaranth Reef
Jun 12, 2025
Sorry, I didn’t noticed that I got logged out before posting the answer
Jack
Jun 19, 2025
No worries. Try lowering the temperature value closer to 0.3.
0.7 is a bit high and can lead to gibberish.
Leave the Top_p as is. If you can share a link to an encrypted chat I can test this further. Thanks.
Blush Sprout
Jun 20, 2025
Hey. I had set it high to prevent repeative answers. I had noticed that in many cases, when I asked to create a list of something, the chat has coppied a bigger part of the first point and had copied exact same phrase to the remaining ones from the list
Jack
Jul 7, 2025
Which model are you seeing this with? Venice Large is the best option for coding
Jack
Jul 7, 2025
You can also try using a system prompt to structure the LLM output further.
https://venice.ai/blog/god-mode-the-power-of-ai-system-prompts
Tomato Hovercraft
Jul 7, 2025
Hello. When I run it without login, there is no visible possibility to change the model and it’s left on the default one.
Besides, the paid plans are too expensve for me :-(
Jack
Jul 7, 2025
Sounds like you are on Venice Simple mode. You can toggle this here:
https://venice.ai/faqs#how-to-toggle-simple-mode
Amaranth Reef
Jul 7, 2025
Actually, the “simple mode” is disabled
Jack
Jul 8, 2025
And you still don’t see any model selection options?
Amaranth Reef
Jul 16, 2025
ok, it seem that it acted up for a bit, but only when I was using it on mobile device.
What about the remaining matters of giving repeative and/or incorrent answers, as well about forgetting the context?
Jack
Jul 22, 2025
Try switching models and please email me [email protected]
So I can take a look at the chats where you are seeing this.
Amaranth Reef
Jul 16, 2025
In my bug report (“Chat act like an idiot”), I had experimented with different settings of P and temperature. Despire of that, I get even worse results like described here…
Jack
Jul 22, 2025
Try switching models and please email me [email protected]
So I can take a look at the chats where you are seeing this.
Amaranth Reef
Jul 22, 2025
Hello. The case is that this model is provably the one (of two avaliable for free plan) that would allow uncensored answers on different questions, aspecially on with other chat’s (aspecially ChatGPT) wouldn’t answer.