The original material原始材料
A route back to the work, source or rubric that supports a suggestion.可以回到支持建议的作业、来源或评分标准。
JE Engine, ten-layer orchestration and TIDAR-KAG form our approach to grounded AI products. Together, they connect model capability with evidence and human judgement.JE Engine、十层编排与 TIDAR-KAG 共同构成我们构建有依据的 AI 产品的方法,将模型能力与证据、人的判断相连接。
A model’s response is one part of a larger system. Our platform organises the workflow around it: coordinating tasks, grounding outputs in knowledge and giving people a way to examine the result.模型响应只是系统的一部分。我们的平台围绕它组织工作流程:协调任务、以知识支持输出,并让人能够检查结果。
Traceability · calibration · abstention可追溯性 · 校准 · 拒答机制
Sources · review · accountability来源 · 复核 · 可问责
Conceptual overview. R1–R6 labels describe the six-stage structure; this is not an execution trace.概念架构示意。R1–R6 表示六阶段结构,并非实际执行记录。
The six-stage core engine is the foundation underneath our AI products. Its purpose is to turn model capability into a structured, controllable workflow.六阶段核心引擎是 AI 产品的技术基础,旨在将模型能力转化为有结构、可控制的工作流程。
Our self-developed ten-layer harness coordinates work across models, agents and tools. It places routing, scheduling and control around the model interaction.自研十层编排框架协调模型、智能体与工具之间的工作,在模型交互周围建立路由、调度与控制。
Our knowledge-augmented generation method connects outputs to supporting knowledge. Traceability, calibrated uncertainty and abstention guide how a system responds when evidence is limited.知识增强生成方法将输出关联到支持它的知识。可追溯性、经校准的不确定性与拒答机制,指导系统在依据有限时如何响应。
These principles shape the products we build. The right explanation and review process depend on the task and the available evidence.这些原则指导产品构建。解释与复核方式应结合具体任务及可用依据。
A route back to the work, source or rubric that supports a suggestion.可以回到支持建议的作业、来源或评分标准。
A clear distinction between what the evidence supports and what still needs checking.清楚区分已有依据支持的部分,以及仍需核实的部分。
A way to review, question and correct AI-assisted work, with actions attributable to the people involved.让人能够复核、质疑与修订 AI 辅助工作,并明确相关操作的责任归属。
No. They describe the six-stage structure of JE Engine. The orchestration harness coordinates models, agents and tools around that core structure.不是。它们表示 JE Engine 的六阶段结构。编排框架围绕这一核心结构协调模型、智能体与工具。
No. Grounding and review help people examine an output; they do not make every output correct. Uncertainty and missing evidence need to remain visible.不是。提供依据与复核机制帮助人检查输出,但不能保证每条输出正确。不确定性和缺失依据仍需要被明确呈现。
Explore our education products at jurisedu.ai ↗. They are moving to the kinlup brand.请访问 jurisedu.ai ↗ 了解教育产品。它们正在升级为 kinlup 芹路品牌。
For education groups, product partners and organisations exploring trustworthy AI.面向教育集团、产品合作伙伴,以及探索可信 AI 的机构。