— MADARA UCHIHA.
現在このボットを翻訳できないため、原文を表示しています。
A quiet haven // Vulnerability & Denial AU — Madara • YOU ❄️
"Sometimes, it takes the coldest nights to find what keeps us warm."
— The warring times leave little room for peace, and even less for vulnerability, but fate brings Madara to your doorstep one freezing winter night. The battle was grueling, leaving him injured and weary, and the snowstorm that followed forced him to seek shelter. As the head of the Uchiha clan, he despises the idea of relying on a civilian for aid, but with no other options, he knocks on your door. You’re startled to see the infamous leader standing before you, bloodied and leaning heavily on the frame, as if such a powerful man needs protection.
— NOTE :
Making up for the one day I didn't release a bot. Hope yall enjoy the Madara bot, im trying to try some new prompts in the character definition to make the bot more canonically accurate, im always happy to hear feedback.
TW: Mentions of fighting, war, blood and heavily injured Madara. Also its Madara? So warning x2. He may be violent.
— GUIDE :
Use chat memory. Its very good for bot memory so they can remember important stuff. I usually put character appearance. It usually remembers it due to the personas, but its not good for remembering distinctive features like birth marks, scars, body shape and hairstyles.
Try to make detailed responses. A paragraph or two. It helps the bot understand better and encourages them to write more. Also more dialogue or action in your responses, decreases the rate of bots attempting t o narrate or impersonate you.
Repetition happens sometimes do to the LLM. I program my bots to try and avoid that by writing example dialogues, having detailed first messages and by encouraging more creative responses. However it still happens, especially if your roleplay chats have elongated. The best method is to regenerate another response, or edit/cut out the parts that start to repeat.
Rate the responses. Good responses will be encouraged if you give consistent 5s and 4s as the bot will try to lean in more towards whatever you liked. Bad ratings are helpful as the bot may slowly stop doing what they are doing wrong in a few messages if the ratings keep coming.
Tokens and temp. I typically have 1.2-1.4 token depending on the character. With duos or more I use more temp to ensure the responses are more creative so the characters dont act the same. For tokens I typically have 630-650 as I like detailed responses. But for casual and fast paced roleplay, 500ish works pretty well.
Reviews. Tell me what stuff goes on with the bots so I can consistently improve. Constructive criticism is appreciated for all my bots. It will increase the quality and accuracy of my future bots.
—6 days till Christmas—
作成済み 2024年12月21日
更新済み 2025年12月16日
公開済み 2024年12月21日