当前位置:首页 > 报告详情

利用知识图谱增强LLM的可解释性和可信度.pdf

上传人: 竿*** 编号:981517 2025-11-29 19页 3.96MB

1、Enhance LLMs explainability and trustworthiness with knowledge graphsLeann Chen GenAI Developer Advocate,DiffbotTL;DR:Knowledge Graphs=Structured Data with ConnectionsExample#1The lost-in-the-middle-phenomenon of LLMsRAG with Hayao Miyazakis wiki page as knowledge baseRAG with Hayao Miyazakis Wiki P

2、age as Knowledge Basehttps:/ effects in LLM-based systemsRAG with Elon Musks wiki page as knowledge baseThe LLM PipelineData+Tokenizing+Embedding Model+Language Model+Chunking Strategy+(Orchestration)+Retrieval Techniques+GenerationHallucination in RAG still?The Choice of Embeddings and Language Mod

3、els matterHard for unstructured data to solve-sorting,filtering,aggregation,etcQuestion:Which article has the lowest sentiment?Question:Whats the latest news?How can structured information help with more accurate information?Back to Example#1Downstream effects in LLM-based systemsRAG with Hayao Miyazakis wiki page as knowledge baseLive Demohttps:/ Demo(images to be replaced with actual live demo)https:/ Demo(images to be replaced with actual live demo)https:/ Demo(images to be replaced with actual live demo)https:/

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。
竿头日上
竿头日上 专家认证

希望事业能一帆风顺,蒸蒸日上!!!

客服
商务合作
小程序
服务号
折叠