https://www.hansking.cn/blog A posts created with Next.js and Tailwind.css en-us admin@hansking.cn (HANS) admin@hansking.cn (HANS) Sun, 08 Dec 2024 10:00:00 GMT https://www.hansking.cn/blog/2026/2024-nian-zong-jie-LLM-ying-yong-gong-cheng-hua-jian-cha-qing-dan https://www.hansking.cn/blog/2026/2024-nian-zong-jie-LLM-ying-yong-gong-cheng-hua-jian-cha-qing-dan [{"value":"2024 年总结：LLM 应用工程化检查清单","url":"#2024-年总结llm-应用工程化检查清单","depth":1},{"value":"一、先确认：你解决的是问题，不是展示模型","url":"#一先确认你解决的是问题不是展示模型","depth":2},{"value":"二、模型选型：不要只看榜单第一名","url":"#二模型选型不要只看榜单第一名","depth":2},{"value":"三、Prompt：从“会说话”走向可维护","url":"#三prompt从会说话走向可维护","depth":2},{"value":"四、结构化输出：让字符串接受类型检查","url":"#四结构化输出让字符串接受类型检查","depth":2},{"value":"五、RAG：真正难的是数据和闭环","url":"#五rag真正难的是数据和闭环","depth":2},{"value":"六、多模态：输入变多，验证不能变少","url":"#六多模态输入变多验证不能变少","depth":2},{"value":"七、长上下文：能放下不等于能用好","url":"#七长上下文能放下不等于能用好","depth":2},{"value":"八、微调：先确认 Prompt 和数据真的不够","url":"#八微调先确认-prompt-和数据真的不够","depth":2},{"value":"九、安全：Prompt Injection 不是模型一句话能解决的","url":"#九安全prompt-injection-不是模型一句话能解决的","depth":2},{"value":"十、可观测性：没有 Trace 就没有复盘","url":"#十可观测性没有-trace-就没有复盘","depth":2},{"value":"十一、成本：把一次回答当成完整账单","url":"#十一成本把一次回答当成完整账单","depth":2},{"value":"十二、发布门禁：不要把生产当实验室","url":"#十二发布门禁不要把生产当实验室","depth":2},{"value":"十三、我在 2024 年留下的工程排序","url":"#十三我在-2024-年留下的工程排序","depth":2},{"value":"十四、我的总结：工程化就是把惊喜变成可重复","url":"#十四我的总结工程化就是把惊喜变成可重复","depth":2}] Sun, 08 Dec 2024 10:00:00 GMT admin@hansking.cn (HANS) AILLM工程实践RAG多模态 https://www.hansking.cn/blog/2026/Claude3-xi-lie-yu-mo-xing-neng-li-ping-ce-fang-fa https://www.hansking.cn/blog/2026/Claude3-xi-lie-yu-mo-xing-neng-li-ping-ce-fang-fa [{"value":"Claude 3 系列与模型能力评测方法","url":"#claude-3-系列与模型能力评测方法","depth":1},{"value":"一、Claude 3 系列为什么值得写进发展史","url":"#一claude-3-系列为什么值得写进发展史","depth":2},{"value":"二、先建立能力矩阵","url":"#二先建立能力矩阵","depth":2},{"value":"三、评测集比一句神奇 Prompt 更重要","url":"#三评测集比一句神奇-prompt-更重要","depth":2},{"value":"四、自动指标和人工指标要分工","url":"#四自动指标和人工指标要分工","depth":2},{"value":"五、不能把长答案当成高质量","url":"#五不能把长答案当成高质量","depth":2},{"value":"六、长上下文要测“找得到”和“用得对”","url":"#六长上下文要测找得到和用得对","depth":2},{"value":"七、多模态评测要看视觉细节","url":"#七多模态评测要看视觉细节","depth":2},{"value":"八、调用参数要固定，结果才可比较","url":"#八调用参数要固定结果才可比较","depth":2},{"value":"九、模型裁判可以用，但不能独裁","url":"#九模型裁判可以用但不能独裁","depth":2},{"value":"十、评测结果要能指导路由","url":"#十评测结果要能指导路由","depth":2},{"value":"十一、上线后继续评测","url":"#十一上线后继续评测","depth":2},{"value":"十二、我的总结：模型评测是地图，不是冠军赛","url":"#十二我的总结模型评测是地图不是冠军赛","depth":2}] Sun, 28 Apr 2024 10:00:00 GMT admin@hansking.cn (HANS) AILLMClaude 3模型评测多模态 https://www.hansking.cn/blog/2026duo-mo-tai-Agent-rang-mo-xing-du-tu-du-biao-du-ping-mu https://www.hansking.cn/blog/2026duo-mo-tai-Agent-rang-mo-xing-du-tu-du-biao-du-ping-mu [{"value":"多模态 Agent：让模型读图、读表、读屏幕","url":"#多模态-agent让模型读图读表读屏幕","depth":1},{"value":"一、先区分三类多模态任务","url":"#一先区分三类多模态任务","depth":2},{"value":"1. 视觉问答","url":"#1-视觉问答","depth":3},{"value":"2. 视觉结构化","url":"#2-视觉结构化","depth":3},{"value":"3. 视觉行动","url":"#3-视觉行动","depth":3},{"value":"二、图像预处理决定模型能不能看清","url":"#二图像预处理决定模型能不能看清","depth":2},{"value":"三、表格不能只当成一张图片","url":"#三表格不能只当成一张图片","depth":2},{"value":"四、结构化输出要有证据和校验","url":"#四结构化输出要有证据和校验","depth":2},{"value":"五、屏幕理解的关键是状态，不是坐标","url":"#五屏幕理解的关键是状态不是坐标","depth":2},{"value":"六、Agent 要知道什么时候自己不确定","url":"#六agent-要知道什么时候自己不确定","depth":2},{"value":"七、评测多模态 Agent 要测整条链路","url":"#七评测多模态-agent-要测整条链路","depth":2},{"value":"八、我最后留下的工程原则","url":"#八我最后留下的工程原则","depth":2}] Tue, 19 May 2026 10:00:00 GMT admin@hansking.cn (HANS) AILLM多模态AgentComputer Usetop