可爱的中文大语言模型训练信息图

提示词

{"type":"可爱教育信息图海报","topic":"{argument name=\"headline text\" default=\"大语言模型的训练过程\"}","subtitle":"{argument name=\"subtitle text\" default=\"从海量数据中学习，变成\"超会聊天\"的小助手！\"}","style":{"overall":"可爱的中文讲解海报，柔和课堂风信息图，圆角分区，温馨奶油色背景，手绘卡通美学，干净的矢量感插画，温暖友好的教育氛围","palette":["奶油色","薰衣草紫","天蓝色","薄荷绿","黄橙色","粉色","柔和棕色"],"rendering":"高质量平涂插画，带轻微阴影，清晰的中文排版，贴纸感装饰，点缀小星星、爱心、箭头和对话气泡"},"layout":{"format":"横版海报","grid":{"rows":2,"columns":4,"count":8},"sections":[{"title":"1. 数据收集","position":"左上","count":1,"labels":["网页","新闻","对话"]},{"title":"2. 数据预处理","position":"顶部一行，第二列","count":1,"labels":["今天 天气 真不错!!","今天|天气|真|不错!!"]},{"title":"3. 预训练","position":"顶部一行，第三列","count":1,"labels":["今天气","很好","不错","?","......"]},{"title":"4. 监督微调（SFT）","position":"右上","count":1,"labels":["问：太阳为什么会发光？","答：因为……","好回答！"]},{"title":"5. 奖励模型训练（RM）","position":"左下","count":1,"labels":["回答A 更好!","回答B 一般般"]},{"title":"6. 强化学习（RLHF）","position":"底部一行，第二列","count":1,"labels":["奖励 +1","惩罚 -1"]},{"title":"7. 评估与测试","position":"底部一行，第三列","count":1,"labels":["知识能力","推理能力","安全性","稳定性"]},{"title":"8. 部署与应用","position":"右下","count":1,"labels":["聊天","写代码","写文章"]}],"topDecorations":{"count":4,"items":["左上角的小闪光","标题周围的粉色彩纸碎屑线条","右上角拿着星星魔法棒的可爱白色吉祥物","写着 冲鸭! 的对话气泡"]},"bottomDecorations":{"count":4,"items":["左下角戴蝴蝶结的小白色吉祥物","带有7个圆角小方框和箭头的总结条","底部居中的结束句","右下角带鼓励文字和爱心的便签"]}},"characters":{"main mascots":{"count":4,"types":["脸颊粉粉的白色胖乎乎熊形吉祥物","带深色面屏和绿色天线的小圆机器人","扎马尾的人类女老师","拿着放大镜的毛茸茸白色角色"]},"recurring_robot_design":"短小可爱的机器人，圆润机身，浅绿色和奶油色外壳，海军蓝面屏带发光眼睛，四肢小巧"},"sectionDetails":[{"title":"1. 数据收集","panelColor":"薰衣草紫","scene":"一只戴着紫色帽子的白色吉祥物坐在笔记本电脑旁，同时身边堆着一摞彩色书本；漂浮的圆角标签代表互联网来源","caption":"数据越多，知识越丰富！"},{"title":"2. 数据预处理","panelColor":"天蓝色","scene":"戴着蓝色帽子的吉祥物把散落的纸屑扫进桶里；一团乱的句子在箭头指引下变成分词后的整洁文本","caption":"把脏乱差，变得整整齐齐～"},{"title":"3. 预训练","panelColor":"薄荷绿","scene":"机器人正在读一本打开的绿色书本，周围的对话气泡里显示着简单的词元和回应，表现语言学习的过程","caption":"像小朋友学习说话一样！"},{"title":"4. 监督微调（SFT）","panelColor":"金黄色","scene":"老师指着一张标注好的问答卡片，而一只可爱的吉祥物认真听讲；重点突出人工标注的高质量问答对","caption":"老师带着学，回答更靠谱！"},{"title":"5. 奖励模型训练（RM）","panelColor":"粉色","scene":"机器人站在绿色对勾和红色叉号之间，通过比较两个回答选项来学习哪个更好","caption":"学会\"挑好答案\"！"},{"title":"6. 强化学习（RLHF）","panelColor":"薰衣草蓝","scene":"机器人拿着工具，奖励板上显示正负分反馈和箭头","caption":"鼓励好的，改掉不好的！"},{"title":"7. 评估与测试","panelColor":"蓝色","scene":"一只毛茸茸的白色检查员角色拿着放大镜，站在带有四个绿色对勾的清单板旁边","caption":"全方位体检，确保质量过关！"},{"title":"8. 部署与应用","panelColor":"柔和红粉色","scene":...