当前位置:首页 > 报告详情

GenAI 提示 Slam:成为 AI 低语者.pdf

上传人: 五**** 编号:1241097 2026-05-16 19页 478.50KB

1、GenAI Prompt Slam:GenAI Prompt Slam:Becoming an AI WhispererBecoming an AI WhispererMarydee Ojala,Editor,Online Searcher(CIL)&Editor,ILI365 eNewsArthur Weiss,Managing Partner,AWAREComputers in Libraries 20261What the heck is a prompt slam?Sort of like a poetry slam Well review some basics of prompti

2、ng and how those change over time and in context Then well move on to some scenarios for which you will create prompts Finally,well try them out There might be prizesComputers in Libraries 20262What we wont do Access proprietary data in an organizations LLM Encourage you to give away“state secrets”H

3、ave the time to fully iterate on prompts Which is unfortunate,since clarifying and refining prompts is essential in the real worldComputers in Libraries 20263A haiku to get us started(Claude wrote this)Ask with clarityPlant seeds of understandingKnowledge blooms and grows.Computers in Libraries 2026

4、4What is prompt engineering?Can refer to two slightly different things 1.We often equate prompt engineering with entering a prompt at a chatbot such as ChatGPT,Copilot,Gemini,Claude.Perplexity,etc.The response from the chatbot is based on the language model that sits behind it May be external or int

5、ernal to our organizations,paid v free versions can differ 2.Web developers and those training the language models view the engineering part of prompt engineering as basic to their jobs Prompt engineering to train models is a behind the scenes activity.Responses to queries are tweaked and modified t

6、o provide more relevant responses but those are not really seen by people entering promptsComputers in Libraries 20265Prompt engineering strengths&weaknesses GenAI is not strong on returning factual information Hallucinations,Just plain gets things wrong(but it look plausible)RAG helps with this,as

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
1. **Prompt Slam定义**:类似诗歌 slam,涵盖提示基础、场景实践及迭代优化,强调提示工程(prompt engineering)的双重含义——用户交互与模型训练。 2. **核心能力**:GenAI擅长创意(总结、写作、分析)但弱于事实准确性(易出现幻觉),RAG和XML提示可缓解。 3. **提示框架**:列举多种结构化方法(如RISE、RTF、STAR)及关键原则(明确步骤、提供示例、迭代调整)。 4. **语言与风险**:术语动态变化可能导致歧义,需注意角色设定、用词准确性(如“senior”的多义性),并警惕AI的假设错误。 5. **实践建议**:提示需包含上下文、约束条件、输出格式及推理要求,跨模型测试可揭示结果差异。
**AI提示技巧?** **如何优化提示?** **提示工程误区?**
客服
商务合作
小程序
服务号
折叠