Page 678 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

scenarios. These capabilities address the traditional challenges of slow information processing
and low resource allocation efficiency in urban management, resulting in a 25% improvement
in urban management efficiency in practical applications and provide scientific decision-making
support for grassroots governance.

The platform fully considers international requirements, supporting multilingual interaction and
cross-cultural adaptability. It also supports RAG to use different city background data. Every city
can establish its own knowledge base,and it allows for customized deployment according to the
urban governance needs of different countries and regions. The platform integrates external
knowledge bases with large language models (LLMs) through RAG, establishing a dynamically
collaborative data ecosystem. Platform administrators can store frequently asked questions,
standardized answers, and relevant policies in the knowledge base in both structured and
vectorized formats. Policy texts and Q&A pairs are transformed into high-dimensional vectors
via embedding models and stored in a vector database for semantic retrieval, while other
structured data is maintained in a relational database. For example, when a user submits a query,
the platform first retrieves matching policy excerpts and case studies through vector similarity
search, then performs logical validation using structured data, ultimately generating accurate
responses and triggering corresponding actions such as automatic work order dispatch. User
feedback and expert review results continuously optimize the knowledge base ranking and
model parameters, ensuring the system maintains high accuracy even after policy updates.h
Due to data privacy concerns, the platform can be privately deployed on government-owned
servers. Alternatively, we offer connectivity to China Unicom cloud servers, which provide high-
level security and trusted data circulation technologies. To ensure the credibility of AI reasoning
and the effectiveness of actions, the platform adopts a closed-loop system combining "expert
oversight + transparency mechanisms + tiered response": First, domain experts manually
review high-risk decisions, and models are continuously refined through feedback. Additionally,
all AI outputs must include explainable justifications (e.g., policy clauses, similar cases). The
platform automatically implements tiered responses based on reasoning confidence: high-
confidence inferences trigger immediate execution, while low-confidence cases are escalated
for human intervention or require supplementary evidence.

Although the initial deployment of the platform may face issues such as high hardware costs,
continuous optimization of algorithms and model scheduling can gradually reduce resource
consumption and expand the scope of application. Meanwhile, with the advancement of
communication technologies and the widespread adoption of 5G/6G, the platform will achieve
lower-latency responses and support the integration of more urban devices.

Use Case Status:

The use case is part of a larger product development.

2�2 Benefits of use case

The Yuanjing Integrated Urban Management Platform, through the application of advanced
large model technology, enhances urban management efficiency and optimizes resource
allocation, providing long-term value for the sustainable development goals of cities and
communities.

642