Original title: Large Language Models (LLMs) for Telecom Root Cause Analysis (RCA): A Structured Reasoning Framework for Evidence-Grounded Diagnosis
Hao Zhou, Mandar Kulkarni, Hao Chen, Yan Xin +2 · September 2026
When cellular networks like 5G slow down or break, engineers need to find the root cause quickly. These modern networks have tangled layers, making diagnosis hard. Standard large language models can reason well, but they often hallucinate false facts, reason unstably, and ignore actual network data. To fix this, researchers designed a structured reasoning framework for language models. It works in three steps. First, it organizes messy network activity data into clean contexts. Next, it forces the model to follow strict decision paths as it diagnoses the problem. Finally, it produces explanations rooted in real network evidence. In tests on two 5G datasets, this method beat baseline tools. It reliably improved diagnostic accuracy and made the system reach more consistent decisions when identifying faults.
You are a mechanic trying to figure out why a car won't start, but someone is only describing the symptoms over the phone. Without being able to open the hood, you end up guessing — and guessing wrong. In this paper, a language model diagnosing network faults faces the same problem: it gets fragments of data and fills in the gaps with guesses. The framework forces the model to organize the information into a standard format, work step by step, and cite specific evidence before declaring what is broken.
The framework helps telecom engineers diagnose network faults more reliably using AI. Modern 5G and 6G networks are layered and complex, and when something goes wrong, tracing the problem back to its original source is hard. Making AI-assisted diagnosis trustworthy means faster fixes, less downtime, and lower operational costs.
The framework was tested only on two 5G datasets (TeleLogs and TelecomTS), so its performance on 6G networks or other contexts is not established. The abstract does not address computational cost or inference latency, which would matter for real-time network operations. The scope of baseline comparisons is not fully detailed, leaving open how the framework performs against the full range of existing techniques.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Large Language Models (LLMs) for Telecom Root Cause Analysis (RCA): A Structured Reasoning Framework for Evidence-Grounded Diagnosis — Hao Zhou, Mandar Kulkarni, Hao Chen et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.