知识卡片
模型代写用户话不是真实输入
模型可能继续生成Human或Parent一方的话,包括擅自写出同意。界面应限制一轮输出并解析角色,状态系统只能根据实际收到的用户事件记录授权;重复一句只回答一轮的提示不是可靠的权限机制。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- · 查看原书依据
Okay, that’s a start! Your bot definitely was able to generate a reasonable response that could be given by a math chatbot. Unfortunately, it also generated a response for the student. You’ll need to tweak your prompt to make sure that doesn’t happen. But the more important question is whether the bot will remember what was said previously in the conversation. Let’s see:
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第12章《对话引擎与会话设计》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-12.html)
- 结论依据:摘录中关于模型代写用户话不是真实输入的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:Okay, that’s a start! Your bot definitely was able to generate a reasonable response that could be given by a math chatbot. Unfortunately, it also generated a response for the student. You’ll need to tweak your prompt to make sure that doesn’t happen. But the more important question is whether the bot will remember what was said previously in the conversation. Let’s see: