Hermes 'telecom brain' advances autonomous mobile networks
Global collaboration yields breakthrough LLM framework for autonomous networks
#UAE #telecom - A global team of researchers collaborating with Abu Dhabi’s Khalifa University has developed a breakthrough multi-agent framework built on large language models (LLMs) designed to autonomously manage and optimise cellular networks. Named Hermes, the new LLM framework uses Network Digital Twins (NDTs) and a ‘blueprint’ approach for autonomous network modeling, allowing it to model and automate diverse network configurations, tasks and behaviours. Overcoming common challenges experienced with LLMs, such as difficulties with network reasoning and computational tasks, Hermes outperforms conventional LLMs, achieving success rates of up to 82.5% for diverse network tasks.
SO WHAT? - Hermes represents a significant step forward in the journey to create fully-autonomous cellular networks, addressing critical operational challenges posed by scale and complexity. LLMs are now thought to be crucial to creating fully-autonomous networks, because of their versatility, being able to make AI agents smarter and faster, interpret a wide range of data types and to communicate with humans. However, there are many technical challenges in using and integrating LLMs for autonomous networks, and a key overall issue is reliability. Hermes uses blueprints to provide structure to its work, allowing the LLM to to reliably model network behavior, and so potentially solving a variety of reliability issues.
Here are some key points regarding Hermes:
Hermes, a new large language model (LLM) framework for autonomous cellular networks has been developed by a global team of researchers from Huawei’s Paris Research Centre, Khalifa University of Science and Technology, the Chinese University of Hong Kong (CUHK) and Yale University. The announcement was made in advance of the 6G Summit Abu Dhabi taking place this week.
The researchers addressed common pitfalls of current LLMs, such as difficulties with network reasoning and computational tasks, to ensure more reliable network operations. Hermes ensures that its model remains reliable and capable of handling a wide range of network tasks.
The new LLM framework uses structured ‘blueprints’ to overcome typical LLM limitations in telecom modeling, employing multiple AI agents for planning, data integration, and precise code generation, marking a breakthrough in using LLMs for specialised domains like telecom.
The structured approach outperforms conventional LLM applications, demonstrating success rates of up to 82.5% in diverse network tasks. Blueprints allow the framework to consistently achieve a 75% success rate in complex tasks such as new base station deployment, significantly outperforming other models like CoT (chain-of-thought), which can only produce a 5% success rate.
Hermes operates as a chain of LLM agents in two roles - Designer and Coder - which iteratively create, test, and refine ‘blueprints’ for each network task, ensuring precise and effective code for various telecom functions.
Network Digital Twins have been implemented in telecom before, but Hermes expands their capability through a structured approach, enabling them to support multiple use cases and configurations more effectively.
Built to handle the dynamic nature of telecom networks, Hermes can adapt autonomously to fluctuations in network load, channel availability, and fault occurrences, maintaining seamless operations.
Automating network management has the potential to lower operational costs significantly, as it eliminates the extensive labor currently required for network oversight and troubleshooting.
The Hermes framework advances research towards the goal of eliminating human oversight in telecom operations.
ZOOM OUT - Autonomous networks are becoming a top priority for telecommunications companies, as they seek to improve performance, reduce costs and provide more seamless subscriber experience. For example, UAE-based operator e& International last month announced that it aims to achieve Level 4 autonomy in its mobile networks by 2030, signing a strategic MoU with Ericsson, Huawei, Intel and Nokia to support this goal.
LINKS
Hermes LLM research paper (arXiv)
Read more about telecom LLMs:
Testing phase launched for TelecomGPT (Middle East AI News)
Abu Dhabi researchers create first-of-its-kind telecom LLM (Middle East AI News)