核心概念
本文介紹了一個名為 Purrfessor 的多模態 AI 聊天機器人,它利用微調的 LLaVA 模型和以人為本的設計原則,提供個性化的飲食指導,並探討了其在提升用戶參與度和滿意度方面的有效性。
統計
αcorrectness = 0.86
αrelevance = 0.96
αclarity = 0.91
αedge case = 0.85
正確率 (M = 7.87)
相關性 (M = 9.4)
清晰度 (M = 9.6)
處理邊緣情況 (M = 9.0)
引用
“The waiting time is long. You can make it same time typing to add interaction.”
“Whenever I ask a question, I have to wait, with my question still in the input box, until the chatbot finishes its response.”
“The version of output can be improved and the answer format like greetings can add more fun language. Emoji to fit the robot personality.”
“The answers seem accurate, useful, and on point; however, when I ask follow-up questions, it does not consider the prior questions I asked.”
“Adapt the recipe based on the user’s preference and hometown.”
“At the beginning, I didn’t know what to do. If the initial page gave me some hints or introductions, I might be clearer.”
“You could add some suggestions for users to start a conversation with the chatbot.”