Page 329 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 8: Intelligent Agent Framework and Practice for Network

Operations and Maintenance Based on the Collaboration of Large
and Small Models 4.3 - 5G

Organization: China Telecom Jiangsu Branch

Country: China

Contact Person(s):

wenhan Rong, rongwh@ chinatelecom �cn, 15301582482

1 Use Case Summary Table

Item Details
Category 5G

Network complexity is increasing, with global failures occurring frequently.
Problem Approximately 70% of failures are attributed to human factors. The lengthy
Addressed fault repair time due to multiparagraph troubleshooting and poor planning
can result in a significant economic loss.

• Cloud-Network operation with strategic manual intervention for precise
fault resolution.
• Deploy digital employees to reduce uncertainties caused by human
Key Aspects of factors.
Solution
• Coordinate large and small models to monitor abnormal events, quickly
establish an emergency war room, and ensure faults are resolved within
10 minutes.

Technology Network fault handling, Reinforcement learning (RL), Knowledge Distilla-
Keywords tion (KD), Agentic AI framework

Data Availability Not Available
Metadata (Type of China Telecom‘s massive network data and knowledge corpus includes
Data) text data, logs, alarms, and configurations.
Using KD technology to automatically generate Chain of Thought (CoT)
Model Training and corpus in the cloud-network operation field, Supervised Fine-Tuning (SFT),
Fine-Tuning and Group Relative Policy Optimization (GRPO) to train cloud-network
operation Large Language Models (LLMs).
Testbeds or Pilot Not Available
Deployments

Code repositories Not Available

293