最新资讯
dev.to 我们在 DolphinBench 上测试记忆检索:纯关键词搜索暴露了自身回归 dev.to 别再把 Agent 写成 while 循环,把它建模为状态机 bloomberg.com Rightmove 与 Scout24 借 AI 产品应对低迷房市并上调展望 stackoverflow.blog 欢迎来到芯片时代:半导体制造、AI 数据中心与芯片设计的未来 www.youtube.com AI Agent 总犯同样的错:CentralAgent AI 将记忆重构为经验系统 pub.towardsai.net Netflix 科学家建议重采样 LLM 评判:5 次仅将误差棒缩小 13% pub.towardsai.net Jev 作为 LLM 前的静默层:构建提示感知的模型路由器 reddit.com OpenAI 称 Astra 仅满足 10% 安全任务标准 dev.to FreeToken:在自有 GPU 上运行前沿 MoE 模型及其内存需求 dev.to 当 AI Agent 说“完成”时,究竟完成了什么? dev.to 部署 Underdog Bark 3.0 Pro:逐字转写、自定义后端与分块设置 dev.to 为艰难对话构建 AI:五个 Agent、一条安全正则与一位盲评监督者
多元动态
登录 注册
智能体 x.com · @petergyang

观点:AI agent 的参与度不等于价值,应看 trace 与 eval

原文 我的收藏

Peter Yang 认为,对 AI agent 而言用户停留时长不代表价值,应通过对话、响应和工具调用等 trace 结合 eval 来评估 agent 是否完成任务。他借此介绍 Amplitude 的 Agent Analytics,并引用其数据:The Economist 达到 96.9% 任务成功率、周任务失败减少 84%;在 2 万+ Amplitude 用户中,首次正向 agent 体验与 3 倍留存相关。上述数据为厂商自述,未经独立验证。

AI推荐理由 提出 agent 场景下“参与度不等于价值”、应看 trace 与 eval 的观点,并给出 Amplitude Agent Analytics 的具体指标(The Economist 任务成功率 96.9%、周任务失败下降 84%;2 万+用户中首次正向 agent 体验与 3 倍留存相关)。但内容带有明显产品推广性质,数据来自厂商自述,缺乏独立验证,需谨慎看待。

62 浏览 3,134 点赞 16 评论 10 转发 2

正文

With AI agents, more user engagement does not necessarily mean more value. A user can spend five minutes talking to an agent and leave frustrated because it never answered their question. Instead, you have to look at the traces - the actual conversations, responses, and tool calls, to understand whether the agent did its job. That’s why I’m excited to try [@Amplitude\_HQ](https://x.com/Amplitude_HQ)'s new Agent Analytics. It combines traces with evals that measure the quality of agent interactions, then connects both to what users do next. This way, teams can see: - What the user was trying to accomplish - Where the agent failed - Whether good and bad answers affected KPIs Early results show why this matters: - The Economist (see below) reached 96.9% task success and cut weekly task failures by 84%. - Across 20K+ Amplitude users, a positive first agent experience was associated with 3× higher retention. Check out Agent Analytics here: [amplitude.com/agent-analytics](https://amplitude.com/agent-analytics) ![Tweet Image](https://pbs.twimg.com/media/HS_qdZpbgAAX2f9.png) ![Amplitude](https://pbs.twimg.com/profile_images/2031507180453605376/uWdX8sJg_normal.png) Amplitude@Amplitude\_HQ Since 1843 [@TheEconomist](https://x.com/TheEconomist) has been fact-checking the world. Now they're fact-checking their AI. Here’s how The Economist improved their AI research assistant with Amplitude Agent Analytics. [bit.ly/4zkd2x0](https://bit.ly/4zkd2x0) [Your browser does not support the video tag.](https://video.twimg.com/amplify_video/2088323272479797249/vid/avc1/1294x720/BzzJqG6dfMrlb8gH.mp4?tag=14) [🔗 View on Twitter](https://x.com/Amplitude_HQ/status/2088323335817941328/video/1) [🔗 View Quoted Tweet](https://x.com/Amplitude_HQ/status/2088323335817941328)

来源报道 · 1 篇 · 以下链接均前往原始媒体

← 返回精选动态 原文

更多内容