Researchers use LLMs to build automated root cause analysis systems that could help telecom networks achieve 99.999% reliability during outages

Study evaluates three LLM approaches—Fine-Tuning, RAG (Retrieval-Augmented Generation), and a Hybrid method—for constructing RCA knowledge bases from support tickets

Testing on real industrial dataset shows generated knowledge bases effectively accelerate root cause analysis tasks in communication networks

Hybrid LLM approach combined with comprehensive lexical and semantic similarity metrics achieved the strongest performance for identifying outage causes