[✅ Solved] unwanted character behaviour

[steep horizon]

I'm using CosmosRP with chub and I can't seem to be able to make character avoid behaviours.
I've tried "{{char}} will never do X", "{{char}} never does X", "{{char}} is strictly forbidden from doing X".
None of these work, and only seem to increase said behaviours.
P.S: I'm not going into details due to the NSFW nature of the exact prompts.

[vernal echo]

When using CosmosRP with ChubAI, please ensure you're using the default generation settings and avoid using any special instructions, jailbreaks, or custom prompts. These often lead to the model producing lower-quality responses or exhibiting unwanted behavior.

Our models are unfiltered, so they can generate a wide range of outputs, including potentially inappropriate content. Here are a few tips to help you:

1. Remove Instructions:

   • If you've used any type of instructions, such as "{{char}} will never do X", you should remove them and start a new chat without custom prompts or jailbreaks. Such instructions can affect the model's behavior adversely.

2. Use Default Settings:

   • Stick with OpenAI's default settings for the best results. For CosmosRP, set the temperature to about 0.85 or as recommended in our Models Channel.

3. Try Different Characters or Websites:

   • If the issue persists, try using a different character or even another platform to determine if it's a platform-specific issue or related to the model itself.

4. NSFW Content:

   • Since the models are unfiltered, be mindful of the possible outputs. Avoid sensitive or NSFW prompts if undesirable behavior is repeatedly a concern.

If the problem persists or if you’re finding the output still doesn't align with what you expect, consider reaching out to the community helpers for additional advice, or if necessary, create a support ticket for more assistance.

Please note: This is an automated response from @vernal echo. While it strive to provide accurate assistance, It may occasionally make mistakes. If you find any inaccuracies or need further clarification, feel free to wait for our community helpers to give you further guidance. Thank you.

[steep horizon]

Ok, so dont write "never do X" into the prompt, but how do I restrict the character then?

[sterile iris]

Try to weave it into the story and in some way try to tell the other character that you don't like it when they do certain things. The bot is pretty smart in figuring out your preferences over the course of the roleplay so just train it that way.

[steep horizon]

That would mean doing that in conversation every time. Not exactly ideal either

[sterile iris]

Just tell the character once and then make note of it in the chat memory. It should refrain from doing whatever you don't want it to do then.

[tame vigil]

Also good to tell the bot how you want it to behave vs how you don't want it to behave

[steep horizon]

By every time I meant every time I start a new chat. I tend to go with the conversation as long as I find it to my tastes and start a new one once I lose interest. So it would be preferrable if I could just have something written in the scenario, description or where ever (preferrably lorebook), and just leave it at that. But reading these, it seems like that isn't really possible. Sorry for the vagueness but again the rules are the rules so.

[vocal night]

Use "(OOC: ???)"

(e.g OOC: the AI shall not make {{char}} do blah blah blah)

(The ai should make {{char}} less agressive)

Things like that

[sterile iris]

Yeah, sorry, there usually isn't something like that. You can't have a permanent solution other than chat memory and guiding the bot through it. The best you can do is make a universal prompt of things you don't want the bot to do and copy paste it from your last chat to the new one

[steep horizon]

Alright, well I tried. Thanks for the help

[vernal echo]

[✅ Solved] unwanted character behaviour