Replies: 2 comments 2 replies
-
|
用哪个函数根据当前情绪、基准情绪和上次修改的时间来确定这次请求前的最新情绪值呢?要允许用户自定义情绪值随时间恢复的函数吗? |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
使用多维向量来表示情绪,而不是单个值,如何呢?情绪值代表的含义有些不太明确 |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment

Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
在 LLM 的返回的 JSON 里添加一个
emotion键,类型为 number。例如返回为
5表示 +5 情绪值,-10表示 -10 情绪值。类似的,在提供给 AI 的提示词里也提示 AI 当前的 累计情绪值,让 AI 根据当前的情绪值自行决定回复的语气。
为了更真实地还原人类情绪变化, Athena 可以根据上次修改情绪值的时间和当前时间的距离调整情绪使其恢复到 平静状态,此时的情绪可以称为 基准情绪。例如上次 AI 和群友大吵一架,过了 24h 情绪逐渐恢复到正常情况。(也就是累计情绪值会趋向于向与基准情绪值的距离减小的方向变化)
Beta Was this translation helpful? Give feedback.
All reactions