AI Risk Management Framework
NIST overview of the voluntary AI Risk Management Framework for incorporating trustworthiness considerations into the design, development, use, and evaluation of AI systems.
GetZed Identity / Knowledge Atlas
游客可以直接浏览、收藏和记录阅读进度,数据先留在当前浏览器。Google 登录只用于云同步和账号管理;登录本身不会上传本机记录,也不代表同意全部条款。
已登录
Google 与 Cloudflare 参与账号及托管处理;跨境、保存期限、导出、删除和未成年人规则见主站法律页。正式公开前仍需专业律师复核。
NIST overview of the voluntary AI Risk Management Framework for incorporating trustworthiness considerations into the design, development, use, and evaluation of AI systems.
OpenAI guidance on designing task-specific evaluations, combining metrics with human judgment, using logs, and treating evaluation as a continuous process.
Internal product method for moving AI-generated ideas through validation, scoring, workbench execution, feedback, and reviewed knowledge cards.
Microsoft HAX guidance that AI products should make it easy for users to edit, refine, undo, or recover when the AI system is wrong.
Official GOV.UK Service Manual guidance on using alpha to prototype, test ideas, and decide which solution is worth taking forward.
Official GOV.UK Service Manual guidance on using discovery to understand users, constraints, costs, and whether a service should move into alpha.
Official GOV.UK Service Manual guidance on improving live services with user research, testing, accessibility, performance metrics, and operational support.
Internal editorial method for turning source metadata, citations, and reviewed paraphrases into reusable Knowledge Atlas cards without publishing private source files.
Google PAIR guidance for human-centered AI product patterns, including feedback loops that let AI systems adapt through user interaction over time.
OpenAI guidance for moving API applications from prototype to production, including security, compliance, testing, misuse prevention, and operational planning.
Official USWDS accessibility guidance that stresses component research, page-level usability testing, and ongoing feedback.
Official USWDS design principles, including starting with real user needs, earning trust, embracing accessibility, and reusing patterns.
Official USWDS search component guidance on input width, recognizable search affordances, and preserving submitted search terms.
Official USWDS validation component guidance on helpful validation feedback and mirroring client-side checks on the server side.
OpenAI guidance that frames eval work as describing a task, running it with test inputs, analyzing results, and iterating on the prompt or system.
一份围绕中国历代王朝兴衰、治理结构、制度选择与历史经验展开的资料,可用于提炼治理周期、组织韧性和制度设计相关知识卡片。
一份介绍中国 34 个省级行政区菜系与代表菜的百科型资料,可用于整理地域饮食文化、菜系分类和地方代表菜知识卡片。