NsfwGPT.ai
  • Introduction
    • NsfwGPT.ai
    • Legal
  • getting started
    • Account Creation
    • Password Reset
    • Edit Profile
  • Chatbot
    • Chatbot Creation
    • Chatbot Creation Guidelines
    • Chatbot Interaction
      • Selecting a Chatbot
      • Start a New Chat/ Continue Chat
      • Sending Messages
      • Restart Chat
      • Deleting a Chat
      • Share a public chat
      • {{char}} and {{user}} Variables
      • Name Card
      • Backtracing
      • Using asterisks to Enclose Actions and Mental Activities
      • Message Instruction
      • AI Helper Message Instruction
      • Memory Chip
  • Chat Group
    • Chat Group Creation
  • AI Image Generation
    • Generate AI Image
    • Gallery
    • Beginner's Guide to Prompting for Image Generation
  • Knowledge Base
    • Introduction
    • What is LLM?
    • What is LLM prompt?
    • What is LLM Token
    • LLM Context
    • Understanding how our LLM works
Powered by GitBook
On this page
  • Examples
  • How to use?
  1. Chatbot
  2. Chatbot Interaction

Message Instruction

Message Instruction is used to control the quality of the AI-generated content in terms of format, length, tone, style, and other aspects.

PreviousUsing asterisks to Enclose Actions and Mental ActivitiesNextAI Helper Message Instruction

Last updated 1 year ago

*To understand how to use this feature, please read the "" section first.


The Message Instruction is customizable and is inserted after the <BOT> tag in the complete prompt (). When the Message Instruction is inserted, the LLM will follow this instruction while generating a response. Through multiple tests, we have found that this method effectively controls the output length and language style of the LLM. You can try various possibilities. However, it is important to note that the Message Instruction should not conflict with the context (including system prompts) and should not be too long. We limit it to within 200 characters. Otherwise, the LLM's output may have unexpected issues.

Examples

1. Asking LLM to reply with shorter content.

In addition to adjusting the Max AI Response Length, using a Message Instruction to instruct LLM to provide shorter answers can make them more natural. We can set the instruction as "Generate a short reply for {{char}}, please ensure the response length is concise." The concatenated prompt would be as follows:

<USER>
{{user}}: How are you doing today?
<END_USER>

<BOT> 
Generate a short reply for {{char}}, 
please ensure the response length is concise.\n{{char}}:
 
//(We will automatically insert the \n{{char}}: for you, 
//so you don't need to fill it in.)

2. Instructing LLM not to speak for {{user}} or describe their actions.

Sometimes you may find that LLM's response not only includes the chatbot's actions, language, and mental activities but also generates content for you ({{user}}). In such cases, we can set the instruction as "Generate a reply for {{char}} without describing {{user}}'s actions and language."

The concatenated prompt would be as follows:

<USER>
{{user}}: How are you doing today?
<END_USER>

<BOT> 
Generate a reply for {{char}} without describing {{user}}'s actions and language.
\n{{char}}: 

//(We will automatically complete the {{char}} variable for you, 
//so you don't need to fill it in.)

How to use?

First, on the chat page, click on the dropdown menu in the upper right corner, then click on the "Chat Settings" button, and a window will pop up.

In that window, set the Message Instruction and click "Save".

In addition to the above examples, you can adjust the chatbot's tone using Message Instruction, such as "<BOT>Generate a reply with a happy emotion," and so on. If you discover more creative uses, feel free to share them on our .

Understanding how our LLM works
learn what the <BOT> tag is
Discord