LWM Multi-Task Optimization Challenge 2025
A community benchmark co-organized with ITU to evaluate foundation models on real-world wireless workloads.
I co-organized the first Large Wireless Models (LWM) challenge together with the International Telecommunication Union. The goal was to stress-test foundation models on practical datasets and create a common benchmark for the community.
- Competition design. I authored the tasks, baseline solutions, and evaluation metrics spanning multi-zone beam selection, channel prediction, and dataset similarity estimation.
- Infrastructure. Built a fully containerized submission system, automated grading scripts, and reproducible leaderboards so that teams could iterate quickly.
- Community support. Hosted office hours, published troubleshooting notes, and mentored student teams across five continents.
Sixty-five teams submitted final solutions, yielding new metrics for foundation-model readiness in wireless systems. We open-sourced the materials at https://lwm-wireless.net/challenge and continue to integrate the best ideas back into the LWM roadmap.