Training and test data (public data) is permanently available here.

A possible approach to construct further synthetic data and fine-tune off-the shelf models is described here.

This challenge provides real-world 5G telecommunication drive-test logs paired with engineering parameters from nearby network cells.

Each training example consists of:

1. A question describing a real network performance degradation scenario
2. A drive-test data excerpt (user-plane measurements across time and location)
3. A set of engineering parameters for relevant 5G cells
4. The ground-truth root cause label among eight possible fault categories (C1–C8)

Your task is to build a model that can understand the interaction between radio KPIs, cell configurations, mobility behavior, and coverage/topology - and correctly predict the most likely root cause of low throughput events.